id
stringlengths 1
12
| question
stringlengths 22
566
| golden_answers
dict | data_source
stringclasses 2
values | prompt
listlengths 1
1
| ability
stringclasses 6
values | reward_model
dict | extra_info
dict | metadata
dict |
|---|---|---|---|---|---|---|---|---|
44894
|
After earning a PhD from Harvard, and while a professor on staff there, Cedric Whitman wrote the book, Homer and the Heroic Tradition, which won him a Nobel Peace Prize in literature.
|
{
"evidence": [
"Nobel Peace Prize_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAfter earning a PhD from Harvard, and while a professor on staff there, Cedric Whitman wrote the book, Homer and the Heroic Tradition, which won him a Nobel Peace Prize in literature.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nobel Peace Prize_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3784,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.158",
"value": "?search="
},
{
"operation": "search",
"time": "18.137",
"value": "Cedric Whitman"
},
{
"operation": "Now on",
"time": "19.006",
"value": "Cedric Whitman"
},
{
"operation": "Highlighting",
"time": "21.71",
"value": "Cedric Whitman_sentence_1"
},
{
"operation": "Highlighting",
"time": "21.711",
"value": "Cedric Whitman_title"
},
{
"operation": "search",
"time": "71.085",
"value": "Contains...Homer and the Heroic Tradition"
},
{
"operation": "Now on",
"time": "72.974",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "85.753",
"value": "Cedric Whitman"
},
{
"operation": "Now on",
"time": "86.148",
"value": "Cedric Whitman"
},
{
"operation": "Highlighting",
"time": "87.709",
"value": "Cedric Whitman_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "89.604",
"value": "Cedric Whitman_sentence_1"
},
{
"operation": "search",
"time": "132.698",
"value": "Nobel Peace Prize"
},
{
"operation": "Now on",
"time": "133.88",
"value": "Nobel Peace Prize"
},
{
"operation": "Highlighting",
"time": "144.027",
"value": "Nobel Peace Prize_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "148.321",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "148.416",
"value": "Cedric Whitman"
},
{
"operation": "Highlighting deleted",
"time": "151.466",
"value": "Cedric Whitman_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "151.467",
"value": "Cedric Whitman_title"
},
{
"operation": "finish",
"time": "161.048",
"value": "finish"
}
]
}
|
68105
|
Margaret Ann Loesch is an American television executive and producer who served as Executive Vice President of Hanna-Barbera Productions and President and Chief Executive Officer of Marvel Productions.
|
{
"evidence": [
"Margaret Loesch_sentence_0",
"Margaret Loesch_sentence_7",
"Margaret Loesch_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMargaret Ann Loesch is an American television executive and producer who served as Executive Vice President of Hanna-Barbera Productions and President and Chief Executive Officer of Marvel Productions.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Margaret Loesch_sentence_0",
"Margaret Loesch_sentence_7",
"Margaret Loesch_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3785,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.071",
"value": "?search="
},
{
"operation": "search",
"time": "9.314",
"value": "Margaret Ann Loesch"
},
{
"operation": "Now on",
"time": "9.959",
"value": "Margaret Loesch"
},
{
"operation": "Highlighting",
"time": "14.473",
"value": "Margaret Loesch_sentence_0"
},
{
"operation": "Highlighting",
"time": "32.4",
"value": "Margaret Loesch_sentence_7"
},
{
"operation": "Highlighting",
"time": "39.096",
"value": "Margaret Loesch_sentence_8"
},
{
"operation": "finish",
"time": "49.561",
"value": "finish"
}
]
}
|
exfever_6080
|
Lily Collins is an actress born in the town of Durham in the southwest of London on the A3 trunk road between the capital of Portsmouth.
|
{
"evidence": [
"Lily Collins",
"Guildford"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLily Collins is an actress born in the town of Durham in the southwest of London on the A3 trunk road between the capital of Portsmouth.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lily Collins",
"Guildford"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3786,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1049
|
The Hundred-Foot Journey is a 2016 American comedy-drama film starring an American actor best known for the film Viceroy's House.
|
{
"evidence": [
"The Hundred-Foot Journey (film)",
"Manish Dayal"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Hundred-Foot Journey is a 2016 American comedy-drama film starring an American actor best known for the film Viceroy's House.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Hundred-Foot Journey (film)",
"Manish Dayal"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3787,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1132
|
Susan Denise Atkins was convicted for her participation in eight of the nine murders, including the most notorious series of murders perpetrated by members of the Manson Family during August 8–10, 1969, in Los Angeles, California, United States.
|
{
"evidence": [
"Susan Atkins",
"Tate murders"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSusan Denise Atkins was convicted for her participation in eight of the nine murders, including the most notorious series of murders perpetrated by members of the Manson Family during August 8–10, 1969, in Los Angeles, California, United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Susan Atkins",
"Tate murders"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3788,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
82781
|
No councilor was elected at the 2015 Stockport Metropolitan Borough Council as it was a fallow year.
|
{
"evidence": [
"2015 Stockport Metropolitan Borough Council election_sentence_3",
"2015 Stockport Metropolitan Borough Council election_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNo councilor was elected at the 2015 Stockport Metropolitan Borough Council as it was a fallow year.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"2015 Stockport Metropolitan Borough Council election_sentence_3",
"2015 Stockport Metropolitan Borough Council election_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3789,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.71",
"value": "?search="
},
{
"operation": "search",
"time": "12.052",
"value": "2015 Stockport Metropolitan Borough Council election"
},
{
"operation": "Now on",
"time": "13.361",
"value": "2015 Stockport Metropolitan Borough Council election"
},
{
"operation": "Page search",
"time": "17.761",
"value": "fallow"
},
{
"operation": "Highlighting",
"time": "34.443",
"value": "2015 Stockport Metropolitan Borough Council election_sentence_3"
},
{
"operation": "Highlighting",
"time": "47.025",
"value": "2015 Stockport Metropolitan Borough Council election_sentence_2"
},
{
"operation": "finish",
"time": "98.929",
"value": "finish"
}
]
}
|
41113
|
The Moroccan government attitudes towards homosexuality tend to be in the interests of the protection of the tradition of the country, and never uses the law as a way to police members of the LGBT+ community; thus it has banned books on homosexuality and required schools to teach a conservative curriculum.
|
{
"evidence": [
"LGBT rights in Morocco_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Moroccan government attitudes towards homosexuality tend to be in the interests of the protection of the tradition of the country, and never uses the law as a way to police members of the LGBT+ community; thus it has banned books on homosexuality and required schools to teach a conservative curriculum.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"LGBT rights in Morocco_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3790,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.718",
"value": "?search="
},
{
"operation": "search",
"time": "35.808",
"value": "Moroccan"
},
{
"operation": "Now on",
"time": "36.737",
"value": "Moroccan"
},
{
"operation": "Highlighting",
"time": "42.465",
"value": "Moroccan_item_1_5"
},
{
"operation": "Highlighting",
"time": "42.465",
"value": "Moroccan_title"
},
{
"operation": "Highlighting",
"time": "42.465",
"value": "Moroccan_section_0"
},
{
"operation": "Highlighting deleted",
"time": "43.513",
"value": "Moroccan_item_1_5"
},
{
"operation": "Highlighting deleted",
"time": "43.514",
"value": "Moroccan_title"
},
{
"operation": "Highlighting deleted",
"time": "43.516",
"value": "Moroccan_section_0"
},
{
"operation": "search",
"time": "70.665",
"value": "Moroccan lgbt"
},
{
"operation": "Now on",
"time": "71.678",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "78.16",
"value": "LGBT rights in Morocco"
},
{
"operation": "Now on",
"time": "78.749",
"value": "LGBT rights in Morocco"
},
{
"operation": "Highlighting",
"time": "95.984",
"value": "LGBT rights in Morocco_sentence_4"
},
{
"operation": "finish",
"time": "105.928",
"value": "finish"
}
]
}
|
exfever_5580
|
The Texas Longhorns men's basketball team represents The University of Texas at Austin at the highest level of intercollegiate athletics sanctioned by the National Collegiate Athletic Association.
|
{
"evidence": [
"Texas Longhorns men's basketball",
"NCAA Division I"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Texas Longhorns men's basketball team represents The University of Texas at Austin at the highest level of intercollegiate athletics sanctioned by the National Collegiate Athletic Association.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Texas Longhorns men's basketball",
"NCAA Division I"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3791,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4589
|
Mr. Freeze is a supervillain who first appeared in #121 of an ongoing American comic book series in which the main character, created by Bob Kane and Bill Finger, first appeared in #27 of an American comic book series published by Detective Comics, later shortened to DC Comics.
|
{
"evidence": [
"Mr. Freeze",
"Batman (comic book)",
"Detective Comics"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMr. Freeze is a supervillain who first appeared in #121 of an ongoing American comic book series in which the main character, created by Bob Kane and Bill Finger, first appeared in #27 of an American comic book series published by Detective Comics, later shortened to DC Comics.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mr. Freeze",
"Batman (comic book)",
"Detective Comics"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3792,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
84214
|
Dustin Kidby was the second in their team from Saskatchewan Regina, Saskatchewan for the 2009 Horizon Laser Vision Center Classic
|
{
"evidence": [
"Dustin Kidby_sentence_9"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDustin Kidby was the second in their team from Saskatchewan Regina, Saskatchewan for the 2009 Horizon Laser Vision Center Classic\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Dustin Kidby_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3793,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.793",
"value": "undefined"
},
{
"operation": "search",
"time": "32.491",
"value": "Dustin Kidby"
},
{
"operation": "Now on",
"time": "33.261",
"value": "Dustin Kidby"
},
{
"operation": "Highlighting",
"time": "36.586",
"value": "Dustin Kidby_sentence_9"
},
{
"operation": "finish",
"time": "47.153",
"value": "finish"
}
]
}
|
exfever_2896
|
William Shakespeare is an English singer born in the market town of Warwickshire, situated in the southeast of a major city and metropolitan borough in the West Midlands, England.
|
{
"evidence": [
"William Shakespeare",
"Stratford-upon-Avon",
"Birmingham"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWilliam Shakespeare is an English singer born in the market town of Warwickshire, situated in the southeast of a major city and metropolitan borough in the West Midlands, England.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"William Shakespeare",
"Stratford-upon-Avon",
"Birmingham"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3794,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47350
|
Sex differences in education are a type of sex equality in the education system, that is evident in Afghanistan, which has the biggest difference in terms of mean years of schooling with a 4.1 difference between male (6) and female (1.9).
|
{
"evidence": [
"Sex differences in education_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSex differences in education are a type of sex equality in the education system, that is evident in Afghanistan, which has the biggest difference in terms of mean years of schooling with a 4.1 difference between male (6) and female (1.9).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sex differences in education_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3795,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.205",
"value": "?search="
},
{
"operation": "search",
"time": "23.936",
"value": "Sex differences in education are a type of sex equality in the education system"
},
{
"operation": "Now on",
"time": "26.629",
"value": "undefined"
},
{
"operation": "search",
"time": "37.744",
"value": "sex equality in the education system"
},
{
"operation": "Now on",
"time": "40.177",
"value": "undefined"
},
{
"operation": "search",
"time": "62.388",
"value": "Education in Afghanistan"
},
{
"operation": "Now on",
"time": "63.178",
"value": "Education in Afghanistan"
},
{
"operation": "Page search",
"time": "66.065",
"value": "sex"
},
{
"operation": "page-search-reset",
"time": "75.583",
"value": "page-search-reset"
},
{
"operation": "search",
"time": "138.177",
"value": "Afghanistan"
},
{
"operation": "Now on",
"time": "148.679",
"value": "Afghanistan"
},
{
"operation": "hyperlink",
"time": "158.819",
"value": "7.5 Education Afghanistan_section_38"
},
{
"operation": "search",
"time": "206.515",
"value": "4.1 difference between male (6) and female (1.9)"
},
{
"operation": "Now on",
"time": "215.276",
"value": "undefined"
},
{
"operation": "search",
"time": "240.081",
"value": "Sex differences in education"
},
{
"operation": "Now on",
"time": "241.71",
"value": "Sex differences in education"
},
{
"operation": "Highlighting",
"time": "250.913",
"value": "Sex differences in education_sentence_2"
},
{
"operation": "finish",
"time": "263.514",
"value": "finish"
}
]
}
|
16954
|
Emmarentia Dam, which is the longest dam in the world since 1980, is about 800 meters (2624.67 ft.), and the Johannesburg Botanical Garden, which grew out of a large rose garden, is on the northern part of the shore.
|
{
"evidence": [
"Emmarentia Dam_sentence_13"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nEmmarentia Dam, which is the longest dam in the world since 1980, is about 800 meters (2624.67 ft.), and the Johannesburg Botanical Garden, which grew out of a large rose garden, is on the northern part of the shore.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Emmarentia Dam_sentence_13"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3796,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.052",
"value": "?search="
},
{
"operation": "search",
"time": "12.145",
"value": "Emmarentia Dam"
},
{
"operation": "Now on",
"time": "13.09",
"value": "Emmarentia Dam"
},
{
"operation": "Highlighting",
"time": "40.877",
"value": "Emmarentia Dam_sentence_13"
},
{
"operation": "finish",
"time": "64.601",
"value": "finish"
}
]
}
|
exfever_7076
|
Radioactive is an electronic rock and alternative rock song with elements of electronic dance music that originated in South Korea in the early 2000s.
|
{
"evidence": [
"Radioactive (Imagine Dragons song)",
"Dubstep"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRadioactive is an electronic rock and alternative rock song with elements of electronic dance music that originated in South Korea in the early 2000s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Radioactive (Imagine Dragons song)",
"Dubstep"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3797,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
86304
|
A full face Motorcycle helmet have vents for airflow and are known for their protectiveness, and does not go to the chin (anterior mandible).
|
{
"evidence": [
"Motorcycle helmet_sentence_26"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nA full face Motorcycle helmet have vents for airflow and are known for their protectiveness, and does not go to the chin (anterior mandible).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Motorcycle helmet_sentence_26"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3798,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.02",
"value": "?search="
},
{
"operation": "search",
"time": "20.192",
"value": "Motorcycle helmet"
},
{
"operation": "Now on",
"time": "22.309",
"value": "Motorcycle helmet"
},
{
"operation": "hyperlink",
"time": "37.594",
"value": "3.1 Full face Motorcycle helmet_section_3"
},
{
"operation": "Highlighting",
"time": "46.193",
"value": "Motorcycle helmet_sentence_26"
},
{
"operation": "finish",
"time": "72.118",
"value": "finish"
}
]
}
|
exfever_7753
|
Take Me High is a 1973 British feature film starring an American actor.
|
{
"evidence": [
"Take Me High",
"Anthony Andrews"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTake Me High is a 1973 British feature film starring an American actor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Take Me High",
"Anthony Andrews"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3799,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
300
|
Air Costa had no headquarters and the fleet of aircrafts was unknown.
|
{
"evidence": [
"Air Costa_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAir Costa had no headquarters and the fleet of aircrafts was unknown.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Air Costa_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3800,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "search",
"time": "0",
"value": "Air Costa"
},
{
"operation": "Highlighting",
"time": "12.988",
"value": "Air Costa_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "23.515",
"value": "Air Costa_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "28.63",
"value": "back-button-clicked"
},
{
"operation": "search",
"time": "148.077",
"value": "Air Costa"
},
{
"operation": "Highlighting",
"time": "169.986",
"value": "Air Costa_sentence_0"
},
{
"operation": "finish",
"time": "317.86",
"value": "finish"
}
]
}
|
13505
|
The Ditrigona cirruncata doesn't have spots on its wings.
|
{
"evidence": [
"Ditrigona cirruncata_sentence_6"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Ditrigona cirruncata doesn't have spots on its wings.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ditrigona cirruncata_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3801,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.964",
"value": "?search="
},
{
"operation": "search",
"time": "68.717",
"value": "Ditrigona cirruncata"
},
{
"operation": "Now on",
"time": "69.477",
"value": "Ditrigona cirruncata"
},
{
"operation": "Highlighting",
"time": "76.555",
"value": "Ditrigona cirruncata_sentence_6"
},
{
"operation": "finish",
"time": "86.172",
"value": "finish"
}
]
}
|
23398
|
The Metaphysical Society of America was sued in 1950 by Paul Weiss.
|
{
"evidence": [
"Metaphysical Society of America_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Metaphysical Society of America was sued in 1950 by Paul Weiss.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Metaphysical Society of America_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3802,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.578",
"value": "?search="
},
{
"operation": "search",
"time": "12.041",
"value": "Metaphysical Society of America"
},
{
"operation": "Now on",
"time": "12.958",
"value": "Metaphysical Society of America"
},
{
"operation": "Page search",
"time": "18.32",
"value": "sue"
},
{
"operation": "Page search",
"time": "27.737",
"value": "suit"
},
{
"operation": "Highlighting",
"time": "34.951",
"value": "Metaphysical Society of America_sentence_0"
},
{
"operation": "finish",
"time": "66.427",
"value": "finish"
}
]
}
|
exfever_8506
|
Jeanine Pirro attended University at Buffalo.
|
{
"evidence": [
"Jeanine Pirro",
"District attorney"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJeanine Pirro attended University at Buffalo.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jeanine Pirro",
"District attorney"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3803,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8028
|
"Just the Way You Are" is the debut solo single by Asian singer-songwriter Bruno Mars, who won a Grammy Award recognizing superior vocal performance by a male in the pop category, that also won by an American singer and musician, who is credited as a pioneer and influence by musicians across a range of genres that includes rhythm and blues.
|
{
"evidence": [
"Just the Way You Are (Bruno Mars song)",
"Grammy Award for Best Male Pop Vocal Performance",
"Stevie Wonder"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"Just the Way You Are\" is the debut solo single by Asian singer-songwriter Bruno Mars, who won a Grammy Award recognizing superior vocal performance by a male in the pop category, that also won by an American singer and musician, who is credited as a pioneer and influence by musicians across a range of genres that includes rhythm and blues.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Just the Way You Are (Bruno Mars song)",
"Grammy Award for Best Male Pop Vocal Performance",
"Stevie Wonder"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3804,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
57566
|
At the first Quebec Winter Carnival in 1894, the first race was organised between Lévis and the Port of Quebec; while the Montreal races were held from 1988 to 1992, during the Fête des Neiges de Montréal .
|
{
"evidence": [
"Ice canoeing_sentence_7",
"Ice canoeing_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAt the first Quebec Winter Carnival in 1894, the first race was organised between Lévis and the Port of Quebec; while the Montreal races were held from 1988 to 1992, during the Fête des Neiges de Montréal .\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Ice canoeing_sentence_7",
"Ice canoeing_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3805,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.34",
"value": "?search="
},
{
"operation": "search",
"time": "50.32",
"value": "quebec winter carnival"
},
{
"operation": "Now on",
"time": "52.085",
"value": "undefined"
},
{
"operation": "search",
"time": "112.315",
"value": "1894 quebec winter carnival"
},
{
"operation": "Now on",
"time": "113.851",
"value": "undefined"
},
{
"operation": "search",
"time": "151.001",
"value": "fete des neiges de montreal"
},
{
"operation": "Now on",
"time": "152.263",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "166.247",
"value": "Ice canoeing"
},
{
"operation": "Now on",
"time": "166.901",
"value": "Ice canoeing"
},
{
"operation": "Highlighting",
"time": "182.04",
"value": "Ice canoeing_sentence_7"
},
{
"operation": "Highlighting",
"time": "191.479",
"value": "Ice canoeing_sentence_8"
},
{
"operation": "finish",
"time": "301.564",
"value": "finish"
}
]
}
|
32537
|
Kolar Plateau in south Karnataka in India is a dry zone suitable for sheep husbandry and famous for silk and milk.
|
{
"evidence": [
"Kolar Plateau_sentence_0",
"Kolar Plateau_sentence_2",
"Kolar Plateau_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKolar Plateau in south Karnataka in India is a dry zone suitable for sheep husbandry and famous for silk and milk.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Kolar Plateau_sentence_0",
"Kolar Plateau_sentence_2",
"Kolar Plateau_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3806,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.66",
"value": "?search="
},
{
"operation": "search",
"time": "6.744",
"value": "Kolar Plateau"
},
{
"operation": "Now on",
"time": "7.258",
"value": "Kolar Plateau"
},
{
"operation": "Highlighting",
"time": "10.371",
"value": "Kolar Plateau_sentence_0"
},
{
"operation": "hyperlink",
"time": "15.144",
"value": "Southern India"
},
{
"operation": "Now on",
"time": "17.268",
"value": "South India"
},
{
"operation": "back-button-clicked",
"time": "20.873",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "20.934",
"value": "Kolar Plateau"
},
{
"operation": "Highlighting",
"time": "28.335",
"value": "Kolar Plateau_sentence_2"
},
{
"operation": "Highlighting",
"time": "34.841",
"value": "Kolar Plateau_sentence_3"
},
{
"operation": "finish",
"time": "59.031",
"value": "finish"
}
]
}
|
exfever_6375
|
Death in Paradise is the first crime drama series created by Robert Thorogood.
|
{
"evidence": [
"Death in Paradise (TV series)",
"Ben Miller"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDeath in Paradise is the first crime drama series created by Robert Thorogood.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Death in Paradise (TV series)",
"Ben Miller"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3807,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1083
|
James L. Brooks was hired by CBS programming executive Grant Tinker.
|
{
"evidence": [
"Broadcast News (film)",
"James L. Brooks"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJames L. Brooks was hired by CBS programming executive Grant Tinker.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Broadcast News (film)",
"James L. Brooks"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3808,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
18077
|
The literacy rate of Mithi Chhapri in the Bhopal District of Madhya Pradesh, India is 35% among the women population and 59% among men, based on the 2011 Census.
|
{
"evidence": [
"Mithi Chhapri_sentence_0",
"Mithi Chhapri_sentence_3"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe literacy rate of Mithi Chhapri in the Bhopal District of Madhya Pradesh, India is 35% among the women population and 59% among men, based on the 2011 Census.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mithi Chhapri_sentence_0",
"Mithi Chhapri_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3809,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.538",
"value": "?search="
},
{
"operation": "search",
"time": "22.636",
"value": "The literacy rate of Mithi Chhapri"
},
{
"operation": "Now on",
"time": "25.367",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "27.284",
"value": "Mithi Chhapri"
},
{
"operation": "Now on",
"time": "28.586",
"value": "Mithi Chhapri"
},
{
"operation": "Highlighting",
"time": "44.188",
"value": "Mithi Chhapri_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "52.09",
"value": "Mithi Chhapri_sentence_0"
},
{
"operation": "Highlighting",
"time": "53.905",
"value": "Mithi Chhapri_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "65.103",
"value": "Mithi Chhapri_sentence_3"
},
{
"operation": "Highlighting",
"time": "123.05",
"value": "Mithi Chhapri_sentence_1"
},
{
"operation": "Highlighting",
"time": "123.631",
"value": "Mithi Chhapri_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "124.404",
"value": "Mithi Chhapri_sentence_1"
},
{
"operation": "hyperlink",
"time": "127.845",
"value": "2011 census of India"
},
{
"operation": "Now on",
"time": "131.858",
"value": "2011 Census of India"
},
{
"operation": "Page search",
"time": "143.339",
"value": "Mithi Chhapri"
},
{
"operation": "hyperlink",
"time": "192.542",
"value": "Madhya Pradesh"
},
{
"operation": "Now on",
"time": "196.661",
"value": "Madhya Pradesh"
},
{
"operation": "Page search",
"time": "202.907",
"value": "Mithi Chhapri"
},
{
"operation": "search",
"time": "228.134",
"value": "Mithi Chhapri"
},
{
"operation": "Now on",
"time": "229.728",
"value": "Mithi Chhapri"
},
{
"operation": "Highlighting",
"time": "239.84",
"value": "Mithi Chhapri_sentence_3"
},
{
"operation": "finish",
"time": "300.265",
"value": "finish"
}
]
}
|
245
|
Veronica Pivetti was born in Naples, started her career as a voice actress at eight years old before moving on to variety shows, films, and ultimately in television for some successful series, including Provaci ancora prof.
|
{
"evidence": [
"Veronica Pivetti_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nVeronica Pivetti was born in Naples, started her career as a voice actress at eight years old before moving on to variety shows, films, and ultimately in television for some successful series, including Provaci ancora prof.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Veronica Pivetti_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3810,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "search",
"time": "0",
"value": "Veronica Pivetti"
},
{
"operation": "Highlighting",
"time": "6.362",
"value": "Veronica Pivetti_sentence_1"
},
{
"operation": "finish",
"time": "24.993",
"value": "finish"
}
]
}
|
exfever_7569
|
Bones is the last television series where Charles Patrick Ryan O'Neal appeared.
|
{
"evidence": [
"Ryan O'Neal",
"Love Story (1970 film)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBones is the last television series where Charles Patrick Ryan O'Neal appeared.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ryan O'Neal",
"Love Story (1970 film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3811,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
44120
|
Sheila Ki Jawani is sung by Vishal Dadlani and Sunidhi Chauhan, the latter of which made her career debut at 23 years old.
|
{
"evidence": [
"Sunidhi Chauhan_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSheila Ki Jawani is sung by Vishal Dadlani and Sunidhi Chauhan, the latter of which made her career debut at 23 years old.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sunidhi Chauhan_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3812,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.515",
"value": "?search="
},
{
"operation": "Now on",
"time": "8.754",
"value": "?search="
},
{
"operation": "search",
"time": "17.656",
"value": "Sheila Ki Jawani (song)"
},
{
"operation": "Now on",
"time": "19.046",
"value": "Tees Maar Khan (2010 film)"
},
{
"operation": "hyperlink",
"time": "199.012",
"value": "Sunidhi Chauhan"
},
{
"operation": "Now on",
"time": "201.091",
"value": "Sunidhi Chauhan"
},
{
"operation": "Highlighting",
"time": "250.849",
"value": "Sunidhi Chauhan_sentence_1"
},
{
"operation": "finish",
"time": "265.85",
"value": "finish"
}
]
}
|
exfever_1033
|
Debut studio album by an American singer who was born on December 14, 1992.
|
{
"evidence": [
"Tori Kelly",
"Unbreakable Smile",
"Tori Kelly"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDebut studio album by an American singer who was born on December 14, 1992.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tori Kelly",
"Unbreakable Smile",
"Tori Kelly"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3813,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5807
|
Northwestern University is a public research university in Evanston, in a state in the Midwestern region of the United States.
|
{
"evidence": [
"Northwestern University",
"Illinois"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNorthwestern University is a public research university in Evanston, in a state in the Midwestern region of the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Northwestern University",
"Illinois"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3814,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
84467
|
1996 British Grand Prix race held on 14 July 1996 at Goldstone, Jacques Villeneuve of Canada got the first place with ten points, while Martin Brundle of United Kingdom got the sixth place with only one point.
|
{
"evidence": [
"1996 British Grand Prix_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n1996 British Grand Prix race held on 14 July 1996 at Goldstone, Jacques Villeneuve of Canada got the first place with ten points, while Martin Brundle of United Kingdom got the sixth place with only one point.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1996 British Grand Prix_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3815,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.945",
"value": "?search="
},
{
"operation": "search",
"time": "7.898",
"value": "1996 British Grand Prix"
},
{
"operation": "Now on",
"time": "8.761",
"value": "1996 British Grand Prix"
},
{
"operation": "Highlighting",
"time": "22.063",
"value": "1996 British Grand Prix_sentence_0"
},
{
"operation": "finish",
"time": "45.218",
"value": "finish"
}
]
}
|
76431
|
Per Høybråten a veterinary and a member of the Christian Democratic Party.
|
{
"evidence": [
"Per Høybråten_sentence_8",
"Per Høybråten_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPer Høybråten a veterinary and a member of the Christian Democratic Party.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Per Høybråten_sentence_8",
"Per Høybråten_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3816,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.815",
"value": "?search="
},
{
"operation": "search",
"time": "8.94",
"value": "Per Høybråten"
},
{
"operation": "Now on",
"time": "9.487",
"value": "Per Høybråten"
},
{
"operation": "Highlighting",
"time": "30.405",
"value": "Per Høybråten_sentence_8"
},
{
"operation": "hyperlink",
"time": "31.068",
"value": "veterinary"
},
{
"operation": "Now on",
"time": "31.741",
"value": "Veterinary"
},
{
"operation": "back-button-clicked",
"time": "35.445",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "35.592",
"value": "Per Høybråten"
},
{
"operation": "Highlighting",
"time": "48.026",
"value": "Per Høybråten_sentence_0"
},
{
"operation": "finish",
"time": "57.629",
"value": "finish"
}
]
}
|
64934
|
The professional wrestling pay-per-view event put on by Juggalo Championship Wrestling took place in 2011 and featured nine different matches.
|
{
"evidence": [
"Hatchet Attacks (2011)_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe professional wrestling pay-per-view event put on by Juggalo Championship Wrestling took place in 2011 and featured nine different matches.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Hatchet Attacks (2011)_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3817,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.399",
"value": "?search="
},
{
"operation": "search",
"time": "39.87",
"value": "Juggalo Championship Wrestling took place in 2011"
},
{
"operation": "Now on",
"time": "41.982",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "54.601",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "Now on",
"time": "55.704",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "Highlighting",
"time": "112.446",
"value": "Juggalo Championship Wrestling_sentence_11"
},
{
"operation": "Highlighting",
"time": "112.449",
"value": "Juggalo Championship Wrestling_title"
},
{
"operation": "Highlighting",
"time": "197.571",
"value": "Juggalo Championship Wrestling_sentence_98"
},
{
"operation": "Highlighting",
"time": "197.575",
"value": "Juggalo Championship Wrestling_section_7"
},
{
"operation": "hyperlink",
"time": "225.853",
"value": "Hatchet Attacks"
},
{
"operation": "Now on",
"time": "226.994",
"value": "Hatchet Attacks (2011)"
},
{
"operation": "Highlighting",
"time": "232.227",
"value": "Hatchet Attacks (2011)_sentence_0"
},
{
"operation": "Highlighting",
"time": "272.129",
"value": "Hatchet Attacks (2011)_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "395.869",
"value": "Hatchet Attacks (2011)_sentence_0"
},
{
"operation": "finish",
"time": "418.087",
"value": "finish"
}
]
}
|
exfever_2458
|
Tunisia is the northernmost country in the world's second-largest and second-most populous continent, after the Earth's largest and most populous continent.
|
{
"evidence": [
"Tunisia",
"Africa",
"Asia"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTunisia is the northernmost country in the world's second-largest and second-most populous continent, after the Earth's largest and most populous continent.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tunisia",
"Africa",
"Asia"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3818,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
81257
|
Bordercrossing Berlin ceased publication after three issues, and featured poetry by authors writing in English (a West Germanic language).
|
{
"evidence": [
"Bordercrossing Berlin_sentence_3",
"Bordercrossing Berlin_sentence_4",
"English language_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBordercrossing Berlin ceased publication after three issues, and featured poetry by authors writing in English (a West Germanic language).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Bordercrossing Berlin_sentence_3",
"Bordercrossing Berlin_sentence_4",
"English language_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3819,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.842",
"value": "undefined"
},
{
"operation": "search",
"time": "14.851",
"value": "Bordercrossing Berlin"
},
{
"operation": "Now on",
"time": "16.053",
"value": "Bordercrossing Berlin"
},
{
"operation": "Highlighting",
"time": "35.637",
"value": "Bordercrossing Berlin_sentence_3"
},
{
"operation": "Highlighting",
"time": "46.584",
"value": "Bordercrossing Berlin_sentence_4"
},
{
"operation": "hyperlink",
"time": "56.557",
"value": "English"
},
{
"operation": "Now on",
"time": "58.947",
"value": "English language"
},
{
"operation": "Highlighting",
"time": "62.529",
"value": "English language_sentence_1"
},
{
"operation": "finish",
"time": "99.318",
"value": "finish"
}
]
}
|
35919
|
John Speidell, on his own, helped push toward the abbreviations of trigonometric functions and published a number of work about mathematics, including An Arithmeticall Extraction in 1628.
|
{
"evidence": [
"John Speidell_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJohn Speidell, on his own, helped push toward the abbreviations of trigonometric functions and published a number of work about mathematics, including An Arithmeticall Extraction in 1628.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"John Speidell_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3820,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.561",
"value": "?search="
},
{
"operation": "search",
"time": "27.86",
"value": "John Speidell"
},
{
"operation": "Now on",
"time": "29.172",
"value": "John Speidell"
},
{
"operation": "Page search",
"time": "49.901",
"value": "1628"
},
{
"operation": "Highlighting",
"time": "56.884",
"value": "John Speidell_sentence_8"
},
{
"operation": "Highlighting",
"time": "68.001",
"value": "John Speidell_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "75.988",
"value": "John Speidell_sentence_8"
},
{
"operation": "finish",
"time": "96.212",
"value": "finish"
}
]
}
|
exfever_389
|
Alec Sokolow is a screenwriter who has worked on computer-animated comics written by the screenwriter behind the movie Cheaper by the Dozen.
|
{
"evidence": [
"Alec Sokolow",
"Toy Story",
"Joel Cohen (writer)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAlec Sokolow is a screenwriter who has worked on computer-animated comics written by the screenwriter behind the movie Cheaper by the Dozen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alec Sokolow",
"Toy Story",
"Joel Cohen (writer)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3821,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
40782
|
Dalton was involved in the Spanish–American War, fighting in Cuba during the Santiago Campaign; then from 1907–1909, he was part of the U.K. occupation force during the Occupation of Cuba.
|
{
"evidence": [
"Albert Clayton Dalton_sentence_12"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDalton was involved in the Spanish–American War, fighting in Cuba during the Santiago Campaign; then from 1907–1909, he was part of the U.K. occupation force during the Occupation of Cuba.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Albert Clayton Dalton_sentence_12"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3822,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.133",
"value": "?search="
},
{
"operation": "search",
"time": "42.194",
"value": "Contains...Dalton was involved in the Spanish–American War"
},
{
"operation": "Now on",
"time": "43.817",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "51.053",
"value": "Albert Clayton Dalton"
},
{
"operation": "Now on",
"time": "51.595",
"value": "Albert Clayton Dalton"
},
{
"operation": "Highlighting",
"time": "64.447",
"value": "Albert Clayton Dalton_sentence_11"
},
{
"operation": "Highlighting",
"time": "71.367",
"value": "Albert Clayton Dalton_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "73.007",
"value": "Albert Clayton Dalton_sentence_11"
},
{
"operation": "finish",
"time": "84.657",
"value": "finish"
}
]
}
|
21632
|
Albert Einstein published theories of general relativity and special relativity, both of which described a unified spacetime in Einstein Field Equations.
|
{
"evidence": [
"Albert Einstein_sentence_18",
"Einstein field equations_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAlbert Einstein published theories of general relativity and special relativity, both of which described a unified spacetime in Einstein Field Equations.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Albert Einstein_sentence_18",
"Einstein field equations_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3823,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.486",
"value": "?search="
},
{
"operation": "search",
"time": "11.463",
"value": "Albert Einstein"
},
{
"operation": "Now on",
"time": "13.368",
"value": "Albert Einstein"
},
{
"operation": "Page search",
"time": "27.531",
"value": "general relativity"
},
{
"operation": "Highlighting",
"time": "41.889",
"value": "Albert Einstein_sentence_18"
},
{
"operation": "Page search",
"time": "61.213",
"value": "unified spacetime"
},
{
"operation": "Page search",
"time": "82.606",
"value": "Einstein Field Equations"
},
{
"operation": "hyperlink",
"time": "96.215",
"value": "Einstein field equations"
},
{
"operation": "Now on",
"time": "96.86",
"value": "Einstein field equations"
},
{
"operation": "Page search",
"time": "107.458",
"value": "special"
},
{
"operation": "Page search",
"time": "123.154",
"value": "relativity"
},
{
"operation": "back-button-clicked",
"time": "173.476",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "173.762",
"value": "Albert Einstein"
},
{
"operation": "Page search",
"time": "185.025",
"value": "special relativ"
},
{
"operation": "hyperlink",
"time": "226.679",
"value": "special relativity"
},
{
"operation": "Now on",
"time": "227.94",
"value": "Special relativity"
},
{
"operation": "Page search",
"time": "273.679",
"value": "spacetime"
},
{
"operation": "Page search",
"time": "281.308",
"value": "unified spacetime"
},
{
"operation": "Page search",
"time": "303.139",
"value": "Einstein Field Equations"
},
{
"operation": "Page search",
"time": "305.09",
"value": "Einstein Field Equation"
},
{
"operation": "search",
"time": "410.111",
"value": "Spacetime"
},
{
"operation": "Now on",
"time": "411.611",
"value": "Spacetime"
},
{
"operation": "Page search",
"time": "419.226",
"value": "relativ"
},
{
"operation": "search",
"time": "592.754",
"value": "general relativity and special relativity unified spacetime"
},
{
"operation": "Now on",
"time": "594.618",
"value": "undefined"
},
{
"operation": "search",
"time": "601.729",
"value": "unified spacetime"
},
{
"operation": "Now on",
"time": "602.637",
"value": "undefined"
},
{
"operation": "search",
"time": "633.854",
"value": "Einstein field equations"
},
{
"operation": "Now on",
"time": "634.288",
"value": "Einstein field equations"
},
{
"operation": "Page search",
"time": "639.677",
"value": "unified"
},
{
"operation": "Page search",
"time": "647.67",
"value": "spacetime"
},
{
"operation": "Highlighting",
"time": "677.753",
"value": "Einstein field equations_sentence_1"
},
{
"operation": "Page search",
"time": "681.23",
"value": "relative"
},
{
"operation": "Page search",
"time": "684.55",
"value": "special rela"
},
{
"operation": "Highlighting deleted",
"time": "693.876",
"value": "Einstein field equations_sentence_1"
},
{
"operation": "Highlighting",
"time": "694.358",
"value": "Einstein field equations_sentence_1"
},
{
"operation": "finish",
"time": "702.898",
"value": "finish"
}
]
}
|
exfever_4351
|
The Twelfth Doctor is an incarnation of the Doctor, the protagonist of the program Doctor Who, by the national broadcaster of the United Kingdom, the largest broadcaster in the world by the number of employees.
|
{
"evidence": [
"Twelfth Doctor",
"BBC"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Twelfth Doctor is an incarnation of the Doctor, the protagonist of the program Doctor Who, by the national broadcaster of the United Kingdom, the largest broadcaster in the world by the number of employees.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Twelfth Doctor",
"BBC"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3824,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
86364
|
Adobe After Effects is a digital visual effects, motion graphics, and composting application developed by Adobe Systems whis is headquartered in San Jose, California and is best known for its Adobe Creative Cloud software.
|
{
"evidence": [
"Adobe After Effects_sentence_0",
"Adobe Inc._sentence_2",
"Adobe Inc._sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAdobe After Effects is a digital visual effects, motion graphics, and composting application developed by Adobe Systems whis is headquartered in San Jose, California and is best known for its Adobe Creative Cloud software.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Adobe After Effects_sentence_0",
"Adobe Inc._sentence_2",
"Adobe Inc._sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3825,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.037",
"value": "?search="
},
{
"operation": "search",
"time": "7.882",
"value": "Adobe After Effects"
},
{
"operation": "Now on",
"time": "9.1",
"value": "Adobe After Effects"
},
{
"operation": "Highlighting",
"time": "10.889",
"value": "Adobe After Effects_sentence_0"
},
{
"operation": "hyperlink",
"time": "49.701",
"value": "Adobe Systems"
},
{
"operation": "Now on",
"time": "50.816",
"value": "Adobe Inc."
},
{
"operation": "Highlighting",
"time": "53.902",
"value": "Adobe Inc._sentence_2"
},
{
"operation": "Highlighting",
"time": "59.34",
"value": "Adobe Inc._sentence_3"
},
{
"operation": "finish",
"time": "84.995",
"value": "finish"
}
]
}
|
exfever_7466
|
Susan Denise Atkins was convicted for her participation in eight of the nine murders, including the most notorious series of murders perpetrated by members of the Manson Family during April 8–10, 1969, in Los Angeles, California, United States.
|
{
"evidence": [
"Susan Atkins",
"Tate murders"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSusan Denise Atkins was convicted for her participation in eight of the nine murders, including the most notorious series of murders perpetrated by members of the Manson Family during April 8–10, 1969, in Los Angeles, California, United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Susan Atkins",
"Tate murders"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3826,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6069
|
Christina María Aguilera is an American singer and songwriter who released a song for her 1999 self-titled debut studio album.
|
{
"evidence": [
"Christina Aguilera",
"Come On Over Baby (All I Want Is You)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChristina María Aguilera is an American singer and songwriter who released a song for her 1999 self-titled debut studio album.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Christina Aguilera",
"Come On Over Baby (All I Want Is You)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3827,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1660
|
Sheyi Emmanuel Adebayor is a Togolese professional wrestler who previously played for a professional football club based in Selhurst in the Borough of Croydon, South London, England.
|
{
"evidence": [
"Emmanuel Adebayor",
"Crystal Palace F.C."
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSheyi Emmanuel Adebayor is a Togolese professional wrestler who previously played for a professional football club based in Selhurst in the Borough of Croydon, South London, England.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Emmanuel Adebayor",
"Crystal Palace F.C."
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3828,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2682
|
The phalanx was composed of heavy infantry armed with a pole weapon consisting of shaft wood or may be made of durable material, such as a chemical element with the symbol Fe and atomic number 24.
|
{
"evidence": [
"Phalanx",
"Spear",
"Iron"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe phalanx was composed of heavy infantry armed with a pole weapon consisting of shaft wood or may be made of durable material, such as a chemical element with the symbol Fe and atomic number 24.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Phalanx",
"Spear",
"Iron"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3829,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3050
|
Christina María Aguilera is an Australian singer who rose to stardom with her album that incorporates dance-pop and a subgenre of pop music that is oriented towards preteens and teenagers.
|
{
"evidence": [
"Christina Aguilera",
"Christina Aguilera (album)",
"Teen pop"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChristina María Aguilera is an Australian singer who rose to stardom with her album that incorporates dance-pop and a subgenre of pop music that is oriented towards preteens and teenagers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Christina Aguilera",
"Christina Aguilera (album)",
"Teen pop"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3830,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1012
|
Alec Baldwin played both leading and supporting roles in a variety of films, including a 1988 Australian romantic comedy-drama film directed by Mike Nichols.
|
{
"evidence": [
"Alec Baldwin",
"Working Girl"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAlec Baldwin played both leading and supporting roles in a variety of films, including a 1988 Australian romantic comedy-drama film directed by Mike Nichols.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alec Baldwin",
"Working Girl"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3831,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4629
|
Legion is a 2010 American supernatural action horror film starring an American actress best known for her role as Dr. Addison Montgomery on the ABC television dramas Grey's Anatomy.
|
{
"evidence": [
"Legion (2010 film)",
"Kate Walsh (actress)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLegion is a 2010 American supernatural action horror film starring an American actress best known for her role as Dr. Addison Montgomery on the ABC television dramas Grey's Anatomy.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Legion (2010 film)",
"Kate Walsh (actress)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3832,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
11448
|
Bulkemsbroek is located far to the village of Simpelveld (municipality and a town in the southeastern Netherlands) in the municipality of Simpelveld, Limburg and the village has a population of 50 people.
|
{
"evidence": [
"Bulkemsbroek_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBulkemsbroek is located far to the village of Simpelveld (municipality and a town in the southeastern Netherlands) in the municipality of Simpelveld, Limburg and the village has a population of 50 people.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bulkemsbroek_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3833,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.696",
"value": "?search="
},
{
"operation": "search",
"time": "9.665",
"value": "Bulkemsbroek"
},
{
"operation": "Now on",
"time": "10.395",
"value": "Bulkemsbroek"
},
{
"operation": "Highlighting",
"time": "13.175",
"value": "Bulkemsbroek_sentence_1"
},
{
"operation": "finish",
"time": "31.138",
"value": "finish"
}
]
}
|
exfever_7307
|
Zain Javadd Malik is a British singer who signed a solo recording contract with an American record label currently owned by Sony Music Entertainment, a subsidiary of Sony Corporation of America.
|
{
"evidence": [
"Zayn Malik",
"RCA Records"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nZain Javadd Malik is a British singer who signed a solo recording contract with an American record label currently owned by Sony Music Entertainment, a subsidiary of Sony Corporation of America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Zayn Malik",
"RCA Records"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3834,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
56870
|
After graduating from Princeton University in 1977, Henry Farber joined the Massachusetts Institute of Technology (a private land-grant research university in Cambridge) as an assistant professor of economics; then in 1991, he left MIT in order to return to Princeton University as a professor of economics, before being endowed with the Hughes-Rogers professorship in 1995.
|
{
"evidence": [
"Henry Farber_sentence_7",
"Massachusetts Institute of Technology_sentence_2",
"Henry Farber_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAfter graduating from Princeton University in 1977, Henry Farber joined the Massachusetts Institute of Technology (a private land-grant research university in Cambridge) as an assistant professor of economics; then in 1991, he left MIT in order to return to Princeton University as a professor of economics, before being endowed with the Hughes-Rogers professorship in 1995.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Henry Farber_sentence_7",
"Massachusetts Institute of Technology_sentence_2",
"Henry Farber_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3835,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.677",
"value": "?search="
},
{
"operation": "search",
"time": "11.869",
"value": "Henry Farber"
},
{
"operation": "Now on",
"time": "12.797",
"value": "Henry Farber"
},
{
"operation": "hyperlink",
"time": "35.564",
"value": "Massachusetts Institute of Technology"
},
{
"operation": "Now on",
"time": "40.003",
"value": "Henry Farber"
},
{
"operation": "Highlighting",
"time": "56.154",
"value": "Henry Farber_sentence_7"
},
{
"operation": "hyperlink",
"time": "59.814",
"value": "Massachusetts Institute of Technology"
},
{
"operation": "Now on",
"time": "69.382",
"value": "Massachusetts Institute of Technology"
},
{
"operation": "Highlighting",
"time": "77.371",
"value": "Massachusetts Institute of Technology_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "78.973",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "79.004",
"value": "Henry Farber"
},
{
"operation": "Highlighting",
"time": "83.896",
"value": "Henry Farber_sentence_8"
},
{
"operation": "finish",
"time": "101.5",
"value": "finish"
}
]
}
|
90281
|
Andy Kyriacou joined Saracens on 1 July 2004 and made his debut in Asian competition against Rugby Parma.
|
{
"evidence": [
"Andy Kyriacou_sentence_9"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAndy Kyriacou joined Saracens on 1 July 2004 and made his debut in Asian competition against Rugby Parma.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Andy Kyriacou_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3836,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.868",
"value": "?search="
},
{
"operation": "search",
"time": "4.366",
"value": "Andy Kyriacou"
},
{
"operation": "Now on",
"time": "5.351",
"value": "Andy Kyriacou"
},
{
"operation": "Highlighting",
"time": "11.265",
"value": "Andy Kyriacou_sentence_8"
},
{
"operation": "Highlighting",
"time": "26.523",
"value": "Andy Kyriacou_sentence_9"
},
{
"operation": "finish",
"time": "30.394",
"value": "finish"
}
]
}
|
35198
|
Yelena Isinbayeva has been in thirteen performances between 2004 and 2012, some of which were in Donetsk, Ukraine (also known as Aleksandrovka )
|
{
"evidence": [
"Yelena Isinbayeva_sentence_0",
"Yelena Isinbayeva_sentence_5"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nYelena Isinbayeva has been in thirteen performances between 2004 and 2012, some of which were in Donetsk, Ukraine (also known as Aleksandrovka )\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Yelena Isinbayeva_sentence_0",
"Yelena Isinbayeva_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3837,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.635",
"value": "undefined"
},
{
"operation": "search",
"time": "11.595",
"value": "Yelena Isinbayeva"
},
{
"operation": "Now on",
"time": "12.732",
"value": "Yelena Isinbayeva"
},
{
"operation": "Highlighting",
"time": "53.6",
"value": "Yelena Isinbayeva_sentence_0"
},
{
"operation": "Highlighting",
"time": "169.666",
"value": "Yelena Isinbayeva_sentence_5"
},
{
"operation": "finish",
"time": "181.188",
"value": "finish"
}
]
}
|
16897
|
Bob Wiese is the all-time top scorer of the Michigan Wolverines football team.
|
{
"evidence": [
"Bob Wiese_sentence_9"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBob Wiese is the all-time top scorer of the Michigan Wolverines football team.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bob Wiese_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3838,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.975",
"value": "undefined"
},
{
"operation": "search",
"time": "9.984",
"value": "Bob Wiese"
},
{
"operation": "Now on",
"time": "10.985",
"value": "Bob Wiese"
},
{
"operation": "Page search",
"time": "25.314",
"value": "score"
},
{
"operation": "Highlighting",
"time": "40.233",
"value": "Bob Wiese_sentence_9"
},
{
"operation": "finish",
"time": "55.908",
"value": "finish"
}
]
}
|
47776
|
Ramón Nocedal Romea(born 31 December 1842 in Madrid) was a Spanish Protestant ultraconservative politician and founder of Integrism who set up La Armonia, a Catholic cultural society which aims to confront heterodoxy in education.
|
{
"evidence": [
"Ramón Nocedal Romea_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRamón Nocedal Romea(born 31 December 1842 in Madrid) was a Spanish Protestant ultraconservative politician and founder of Integrism who set up La Armonia, a Catholic cultural society which aims to confront heterodoxy in education.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ramón Nocedal Romea_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3839,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.05",
"value": "?search="
},
{
"operation": "search",
"time": "19.64",
"value": "Ramón Nocedal Romea"
},
{
"operation": "Now on",
"time": "22.236",
"value": "Ramón Nocedal Romea"
},
{
"operation": "Highlighting",
"time": "35.127",
"value": "Ramón Nocedal Romea_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "36.106",
"value": "Ramón Nocedal Romea_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "36.119",
"value": "Ramón Nocedal Romea_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "36.119",
"value": "Ramón Nocedal Romea_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "40.43",
"value": "Ramón Nocedal Romea_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "45.086",
"value": "Ramón Nocedal Romea_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "45.092",
"value": "Ramón Nocedal Romea_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "45.097",
"value": "Ramón Nocedal Romea_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "67.048",
"value": "finish"
}
]
}
|
37530
|
Watford High Streetwards and Watford Junction are two historical highways.
|
{
"evidence": [
"Watford Junction railway station_sentence_0",
"Watford High Street railway station_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWatford High Streetwards and Watford Junction are two historical highways.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Watford Junction railway station_sentence_0",
"Watford High Street railway station_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3840,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.96",
"value": "?search="
},
{
"operation": "search",
"time": "24.429",
"value": "Contains...watford high streetwards"
},
{
"operation": "Now on",
"time": "25.665",
"value": "undefined"
},
{
"operation": "search",
"time": "36.422",
"value": "Watford Junction railway station"
},
{
"operation": "Now on",
"time": "38.035",
"value": "Watford Junction railway station"
},
{
"operation": "Highlighting",
"time": "45.574",
"value": "Watford Junction railway station_sentence_0"
},
{
"operation": "search",
"time": "60.926",
"value": "Watford High Street railway station"
},
{
"operation": "Now on",
"time": "62.294",
"value": "Watford High Street railway station"
},
{
"operation": "Highlighting",
"time": "66.703",
"value": "Watford High Street railway station_sentence_3"
},
{
"operation": "finish",
"time": "75.985",
"value": "finish"
}
]
}
|
exfever_2710
|
Russell Crowe starred in a film that tells the story of the exploits and downfall of a neo-Nazi group in the blue-collar suburban most populous city in the Australian state of Victoria.
|
{
"evidence": [
"Russell Crowe",
"Romper Stomper",
"Melbourne"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRussell Crowe starred in a film that tells the story of the exploits and downfall of a neo-Nazi group in the blue-collar suburban most populous city in the Australian state of Victoria.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Russell Crowe",
"Romper Stomper",
"Melbourne"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3841,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
26940
|
Institute for Contemporary Culture, which has hosted very few travelling exhibitions, was established in 1989 and is currently directed by Francisco Alvarez.
|
{
"evidence": [
"Institute for Contemporary Culture_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nInstitute for Contemporary Culture, which has hosted very few travelling exhibitions, was established in 1989 and is currently directed by Francisco Alvarez.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Institute for Contemporary Culture_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3842,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.471",
"value": "?search="
},
{
"operation": "search",
"time": "24.428",
"value": "Institute for Contemporary Culture"
},
{
"operation": "Now on",
"time": "25.459",
"value": "Institute for Contemporary Culture"
},
{
"operation": "Highlighting",
"time": "101.407",
"value": "Institute for Contemporary Culture_sentence_3"
},
{
"operation": "finish",
"time": "112.591",
"value": "finish"
}
]
}
|
94372
|
In April 1521, Ferdinand Magellan accompanied Duarte de Menezes from Lisbon to India as the next governor of Portuguese India, a colonial state of the Portuguese Empire founded six years after discovering a sea route to the Indian Subcontinent by the Kingdom of Portugal.
|
{
"evidence": [
"Duarte de Menezes_sentence_8",
"Duarte de Menezes_sentence_7",
"Duarte de Menezes_sentence_6",
"Portuguese India_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn April 1521, Ferdinand Magellan accompanied Duarte de Menezes from Lisbon to India as the next governor of Portuguese India, a colonial state of the Portuguese Empire founded six years after discovering a sea route to the Indian Subcontinent by the Kingdom of Portugal.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Duarte de Menezes_sentence_8",
"Duarte de Menezes_sentence_7",
"Duarte de Menezes_sentence_6",
"Portuguese India_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3843,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.492",
"value": "undefined"
},
{
"operation": "search",
"time": "13.195",
"value": "Ferdinand Magellan accompanied Duarte de Menezes from Lisbon to India as the next governor of Portuguese India, a colonial state of the Portuguese Empire founded six years after discovering a sea route to the Indian Subcontinent by the Kingdom of Portugal."
},
{
"operation": "Now on",
"time": "16.802",
"value": "undefined"
},
{
"operation": "search",
"time": "64.542",
"value": "Ferdinand Magellan accompanied Duarte de Menezes from Lisbon to India"
},
{
"operation": "Now on",
"time": "65.346",
"value": "undefined"
},
{
"operation": "search",
"time": "88.017",
"value": "Duarte de Menezes"
},
{
"operation": "Now on",
"time": "88.868",
"value": "Duarte de Menezes"
},
{
"operation": "Page search",
"time": "98.076",
"value": "ferdinand"
},
{
"operation": "Page search",
"time": "137.877",
"value": "governor"
},
{
"operation": "Page search",
"time": "161.288",
"value": "1521"
},
{
"operation": "Highlighting",
"time": "177.349",
"value": "Duarte de Menezes_sentence_7"
},
{
"operation": "Highlighting",
"time": "178.617",
"value": "Duarte de Menezes_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "181.321",
"value": "Duarte de Menezes_sentence_7"
},
{
"operation": "Highlighting",
"time": "185.173",
"value": "Duarte de Menezes_sentence_7"
},
{
"operation": "Page search",
"time": "203.221",
"value": "governor"
},
{
"operation": "Highlighting",
"time": "212.281",
"value": "Duarte de Menezes_sentence_6"
},
{
"operation": "hyperlink",
"time": "214.265",
"value": "Portuguese India"
},
{
"operation": "Now on",
"time": "214.998",
"value": "Portuguese India"
},
{
"operation": "Page search",
"time": "228.274",
"value": "colonial"
},
{
"operation": "Highlighting",
"time": "230.538",
"value": "Portuguese India_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "238.281",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "238.37",
"value": "Duarte de Menezes"
},
{
"operation": "finish",
"time": "264.228",
"value": "finish"
}
]
}
|
17128
|
Hellboy Omnibus volumes 1-4 are editions that collect the partial Hellboy series in chronological order from May 9, 2018 to September 5, 2018.
|
{
"evidence": [
"Hellboy_sentence_117"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHellboy Omnibus volumes 1-4 are editions that collect the partial Hellboy series in chronological order from May 9, 2018 to September 5, 2018.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Hellboy_sentence_117"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3844,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.557",
"value": "?search="
},
{
"operation": "search",
"time": "16.208",
"value": "Hellboy Omnibus"
},
{
"operation": "Now on",
"time": "17.356",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "41.476",
"value": "Hellboy"
},
{
"operation": "Now on",
"time": "42.598",
"value": "Hellboy"
},
{
"operation": "Page search",
"time": "61.157",
"value": "omnibus"
},
{
"operation": "Page search",
"time": "63.192",
"value": "omnibus"
},
{
"operation": "Highlighting",
"time": "106.448",
"value": "Hellboy_header_cell_7_1_0"
},
{
"operation": "Highlighting",
"time": "107.74",
"value": "Hellboy_header_cell_7_0_0"
},
{
"operation": "Highlighting",
"time": "108.796",
"value": "Hellboy_header_cell_7_2_0"
},
{
"operation": "Highlighting",
"time": "109.456",
"value": "Hellboy_header_cell_7_3_0"
},
{
"operation": "Highlighting",
"time": "110.032",
"value": "Hellboy_header_cell_7_4_0"
},
{
"operation": "Highlighting",
"time": "112.149",
"value": "Hellboy_cell_7_1_2"
},
{
"operation": "Highlighting",
"time": "112.16",
"value": "Hellboy_header_cell_7_1_0"
},
{
"operation": "Highlighting",
"time": "112.161",
"value": "Hellboy_header_cell_7_0_2"
},
{
"operation": "Highlighting",
"time": "115.528",
"value": "Hellboy_cell_7_2_2"
},
{
"operation": "Highlighting",
"time": "115.537",
"value": "Hellboy_header_cell_7_2_0"
},
{
"operation": "Highlighting",
"time": "116.104",
"value": "Hellboy_cell_7_3_2"
},
{
"operation": "Highlighting",
"time": "116.115",
"value": "Hellboy_header_cell_7_3_0"
},
{
"operation": "Highlighting",
"time": "116.92",
"value": "Hellboy_cell_7_4_2"
},
{
"operation": "Highlighting",
"time": "116.93",
"value": "Hellboy_header_cell_7_4_0"
},
{
"operation": "Highlighting",
"time": "124.868",
"value": "Hellboy_sentence_117"
},
{
"operation": "Highlighting deleted",
"time": "135.36",
"value": "Hellboy_header_cell_7_0_0"
},
{
"operation": "Highlighting deleted",
"time": "139.72",
"value": "Hellboy_cell_7_4_2"
},
{
"operation": "Highlighting deleted",
"time": "139.724",
"value": "Hellboy_header_cell_7_4_0"
},
{
"operation": "Highlighting deleted",
"time": "140.408",
"value": "Hellboy_cell_7_3_2"
},
{
"operation": "Highlighting deleted",
"time": "140.411",
"value": "Hellboy_header_cell_7_3_0"
},
{
"operation": "Highlighting",
"time": "140.984",
"value": "Hellboy_cell_7_2_3"
},
{
"operation": "Highlighting",
"time": "140.988",
"value": "Hellboy_header_cell_7_0_3"
},
{
"operation": "Highlighting deleted",
"time": "142.492",
"value": "Hellboy_cell_7_2_3"
},
{
"operation": "Highlighting deleted",
"time": "142.496",
"value": "Hellboy_header_cell_7_0_3"
},
{
"operation": "Highlighting deleted",
"time": "143.132",
"value": "Hellboy_cell_7_2_2"
},
{
"operation": "Highlighting deleted",
"time": "143.137",
"value": "Hellboy_header_cell_7_2_0"
},
{
"operation": "Highlighting deleted",
"time": "143.836",
"value": "Hellboy_cell_7_1_2"
},
{
"operation": "Highlighting deleted",
"time": "143.84",
"value": "Hellboy_header_cell_7_1_0"
},
{
"operation": "Highlighting deleted",
"time": "143.841",
"value": "Hellboy_header_cell_7_0_2"
},
{
"operation": "finish",
"time": "167.996",
"value": "finish"
}
]
}
|
exfever_1789
|
Tom Felton is an English actor best known for playing in the film adaptations of the Harry Potter fantasy novels by J. K. Rowling as a fictional character frequently accompanied by his two cronies, who act as a loyal employee, supporter, or aide to some powerful figure engaged in nefarious or criminal enterprises.
|
{
"evidence": [
"Tom Felton",
"Draco Malfoy",
"Henchman"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTom Felton is an English actor best known for playing in the film adaptations of the Harry Potter fantasy novels by J. K. Rowling as a fictional character frequently accompanied by his two cronies, who act as a loyal employee, supporter, or aide to some powerful figure engaged in nefarious or criminal enterprises.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tom Felton",
"Draco Malfoy",
"Henchman"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3845,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
83540
|
Art Hirahara is a pianist and composer who studied under a famous pianist at Oberlin.
|
{
"evidence": [
"Art Hirahara_sentence_0",
"Art Hirahara_sentence_3"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nArt Hirahara is a pianist and composer who studied under a famous pianist at Oberlin.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Art Hirahara_sentence_0",
"Art Hirahara_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3846,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.262",
"value": "?search="
},
{
"operation": "search",
"time": "77.687",
"value": "Art Hirahara"
},
{
"operation": "Now on",
"time": "78.269",
"value": "Art Hirahara"
},
{
"operation": "Highlighting",
"time": "85.081",
"value": "Art Hirahara_sentence_0"
},
{
"operation": "Highlighting",
"time": "100.067",
"value": "Art Hirahara_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "123.44",
"value": "Art Hirahara_sentence_3"
},
{
"operation": "Highlighting",
"time": "125.58",
"value": "Art Hirahara_sentence_3"
},
{
"operation": "search",
"time": "131.309",
"value": "Neal Creque"
},
{
"operation": "Now on",
"time": "132.246",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "138.038",
"value": "2000s in jazz"
},
{
"operation": "Now on",
"time": "138.898",
"value": "2000s in jazz"
},
{
"operation": "Page search",
"time": "141.701",
"value": "Neal Creque"
},
{
"operation": "Page search",
"time": "142.599",
"value": "Neal Creque"
},
{
"operation": "back-button-clicked",
"time": "150.979",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "151.488",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "153.57",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "153.574",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "153.632",
"value": "Art Hirahara"
},
{
"operation": "Page search",
"time": "161.123",
"value": "famous"
},
{
"operation": "finish",
"time": "196.491",
"value": "finish"
}
]
}
|
exfever_3321
|
Boston Legal is an American legal action-drama television series created by David E. Kelley and starring a Canadian actor best known for his role as Captain of the USS Enterprise in the Star Trek franchise, who leads his crew as they explore new worlds, new civilizations, and boldly go where no man has gone before.
|
{
"evidence": [
"Boston Legal",
"William Shatner",
"James T. Kirk"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBoston Legal is an American legal action-drama television series created by David E. Kelley and starring a Canadian actor best known for his role as Captain of the USS Enterprise in the Star Trek franchise, who leads his crew as they explore new worlds, new civilizations, and boldly go where no man has gone before.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Boston Legal",
"William Shatner",
"James T. Kirk"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3847,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
74196
|
Anker was a member of the Grand Council of Bern from 1870 to 1874, after he was awarded a gold medal at the Paris Salon for Schlafendes Mädchen im Walde in 1866.
|
{
"evidence": [
"Albert Anker_sentence_10",
"Albert Anker_sentence_9"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAnker was a member of the Grand Council of Bern from 1870 to 1874, after he was awarded a gold medal at the Paris Salon for Schlafendes Mädchen im Walde in 1866.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Albert Anker_sentence_10",
"Albert Anker_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3848,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.954",
"value": "?search="
},
{
"operation": "search",
"time": "23.386",
"value": "Anker"
},
{
"operation": "Now on",
"time": "24.077",
"value": "Anker"
},
{
"operation": "search",
"time": "40.324",
"value": "Grand Council of Bern"
},
{
"operation": "Now on",
"time": "41.079",
"value": "Grand Council of Bern"
},
{
"operation": "Page search",
"time": "46.796",
"value": "anker"
},
{
"operation": "search",
"time": "86.293",
"value": "Anker grand council"
},
{
"operation": "Now on",
"time": "87.63",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "95.652",
"value": "Yehudi Menuhin"
},
{
"operation": "Now on",
"time": "96.314",
"value": "Yehudi Menuhin"
},
{
"operation": "search",
"time": "103.335",
"value": "Albert Anker"
},
{
"operation": "Now on",
"time": "104.011",
"value": "Albert Anker"
},
{
"operation": "Page search",
"time": "118.502",
"value": "council"
},
{
"operation": "Highlighting",
"time": "121.751",
"value": "Albert Anker_sentence_10"
},
{
"operation": "Highlighting",
"time": "133.622",
"value": "Albert Anker_sentence_9"
},
{
"operation": "finish",
"time": "181.159",
"value": "finish"
}
]
}
|
28141
|
Robert Scott Moncrieff, a Scottish advocate, amateur illustrator and caricaturist who died in Edinburgh on 18 June 1869, is buried with other members of the Scott Moncrieff family at the Greyfriars Kirkyard.
|
{
"evidence": [
"Robert Scott Moncrieff_sentence_0",
"Robert Scott Moncrieff_sentence_6",
"Robert Scott Moncrieff_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRobert Scott Moncrieff, a Scottish advocate, amateur illustrator and caricaturist who died in Edinburgh on 18 June 1869, is buried with other members of the Scott Moncrieff family at the Greyfriars Kirkyard.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Robert Scott Moncrieff_sentence_0",
"Robert Scott Moncrieff_sentence_6",
"Robert Scott Moncrieff_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3849,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.142",
"value": "?search="
},
{
"operation": "search",
"time": "5.328",
"value": "Robert Scott Moncrieff"
},
{
"operation": "Now on",
"time": "5.919",
"value": "Robert Scott Moncrieff"
},
{
"operation": "Highlighting",
"time": "7.356",
"value": "Robert Scott Moncrieff_sentence_0"
},
{
"operation": "Highlighting",
"time": "17.453",
"value": "Robert Scott Moncrieff_sentence_6"
},
{
"operation": "Highlighting",
"time": "24.057",
"value": "Robert Scott Moncrieff_sentence_7"
},
{
"operation": "finish",
"time": "39.878",
"value": "finish"
}
]
}
|
82752
|
CCR1, the role of which, according to studies, is to protect humans from inflammatory response, and susceptibility to virus and parasite, encodes a member of the beta chemokine receptor family (found on the surface of certain cells that interact with a type of cytokine called a chemokine).
|
{
"evidence": [
"CCR1_sentence_5",
"CCR1_sentence_2",
"Chemokine receptor_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCCR1, the role of which, according to studies, is to protect humans from inflammatory response, and susceptibility to virus and parasite, encodes a member of the beta chemokine receptor family (found on the surface of certain cells that interact with a type of cytokine called a chemokine).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"CCR1_sentence_5",
"CCR1_sentence_2",
"Chemokine receptor_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3850,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.334",
"value": "?search="
},
{
"operation": "search",
"time": "12.431",
"value": "CCR1"
},
{
"operation": "Now on",
"time": "13.482",
"value": "CCR1"
},
{
"operation": "Highlighting",
"time": "34.262",
"value": "CCR1_sentence_5"
},
{
"operation": "Highlighting",
"time": "42.454",
"value": "CCR1_sentence_2"
},
{
"operation": "hyperlink",
"time": "44.09",
"value": "chemokine receptor"
},
{
"operation": "Now on",
"time": "45.098",
"value": "Chemokine receptor"
},
{
"operation": "Highlighting",
"time": "48.637",
"value": "Chemokine receptor_sentence_0"
},
{
"operation": "finish",
"time": "54.427",
"value": "finish"
}
]
}
|
exfever_4784
|
Jennifer Joanna Aniston was the most famous American actress and producer.
|
{
"evidence": [
"Jennifer Aniston",
"Mac and Me"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJennifer Joanna Aniston was the most famous American actress and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jennifer Aniston",
"Mac and Me"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3851,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
50968
|
China's Zhang Nan and Zhao Yunlei ranked 1st in the qualifying round of the mixed doubles event of the 2016 Summer Olympics which was held May 4, 2015 and May 1, 2016.
|
{
"evidence": [
"Badminton at the 2016 Summer Olympics – Qualification_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChina's Zhang Nan and Zhao Yunlei ranked 1st in the qualifying round of the mixed doubles event of the 2016 Summer Olympics which was held May 4, 2015 and May 1, 2016.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Badminton at the 2016 Summer Olympics – Qualification_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3852,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.23",
"value": "?search="
},
{
"operation": "search",
"time": "21.811",
"value": "2016 Summer Olympics"
},
{
"operation": "Now on",
"time": "26.181",
"value": "2016 Summer Olympics"
},
{
"operation": "hyperlink",
"time": "55.555",
"value": "3 The Games 2016 Summer Olympics_section_11"
},
{
"operation": "search",
"time": "117.237",
"value": "China at the 2016 Summer Olympics"
},
{
"operation": "Now on",
"time": "177.002",
"value": "China at the 2016 Summer Olympics"
},
{
"operation": "hyperlink",
"time": "297.88",
"value": "Tennis at the 2016 Summer Olympics"
},
{
"operation": "Now on",
"time": "300.251",
"value": "Tennis at the 2016 Summer Olympics"
},
{
"operation": "hyperlink",
"time": "305.024",
"value": "2 Qualification Tennis at the 2016 Summer Olympics_section_1"
},
{
"operation": "hyperlink",
"time": "326.326",
"value": "Tennis at the 2016 Summer Olympics – Mixed Doubles"
},
{
"operation": "Now on",
"time": "327.523",
"value": "Tennis at the 2016 Summer Olympics – Mixed doubles"
},
{
"operation": "back-button-clicked",
"time": "354.127",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "354.198",
"value": "Tennis at the 2016 Summer Olympics#Qualification Tennis at the 2016 Summer Olympics section 1"
},
{
"operation": "hyperlink",
"time": "433.512",
"value": "Tennis at the 2016 Summer Olympics – Qualification"
},
{
"operation": "Now on",
"time": "440.166",
"value": "Tennis at the 2016 Summer Olympics – Qualification"
},
{
"operation": "search",
"time": "573.066",
"value": "insource"
},
{
"operation": "Now on",
"time": "575.547",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "583.531",
"value": "Competition format Badminton at the 2016 Summer Olympics – Mixed doubles_section_0"
},
{
"operation": "Now on",
"time": "584.879",
"value": "Badminton at the 2016 Summer Olympics – Mixed doubles#Competition format Badminton at the 2016 Summer Olympics .E2.80.93 Mixed doubles section 0"
},
{
"operation": "hyperlink",
"time": "610.581",
"value": "Badminton at the 2016 Summer Olympics"
},
{
"operation": "Now on",
"time": "612.2",
"value": "Badminton at the 2016 Summer Olympics"
},
{
"operation": "hyperlink",
"time": "638.31",
"value": "5.5 Mixed doubles Badminton at the 2016 Summer Olympics_section_13"
},
{
"operation": "hyperlink",
"time": "641.413",
"value": "Badminton at the 2016 Summer Olympics – Mixed doubles"
},
{
"operation": "Now on",
"time": "642.367",
"value": "Badminton at the 2016 Summer Olympics – Mixed doubles"
},
{
"operation": "back-button-clicked",
"time": "648.626",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "648.695",
"value": "Badminton at the 2016 Summer Olympics#Mixed doubles Badminton at the 2016 Summer Olympics section 13"
},
{
"operation": "Highlighting",
"time": "664.207",
"value": "Badminton at the 2016 Summer Olympics_sentence_6"
},
{
"operation": "Highlighting",
"time": "664.208",
"value": "Badminton at the 2016 Summer Olympics_title"
},
{
"operation": "Highlighting",
"time": "664.209",
"value": "Badminton at the 2016 Summer Olympics_section_0"
},
{
"operation": "hyperlink",
"time": "664.846",
"value": "Badminton at the 2016 Summer Olympics – Qualification"
},
{
"operation": "Now on",
"time": "667.558",
"value": "Badminton at the 2016 Summer Olympics – Qualification"
},
{
"operation": "Highlighting",
"time": "685.175",
"value": "Badminton at the 2016 Summer Olympics – Qualification_cell_5_1_2"
},
{
"operation": "Highlighting",
"time": "685.177",
"value": "Badminton at the 2016 Summer Olympics – Qualification_section_2"
},
{
"operation": "Highlighting",
"time": "685.177",
"value": "Badminton at the 2016 Summer Olympics – Qualification_section_7"
},
{
"operation": "Highlighting",
"time": "685.178",
"value": "Badminton at the 2016 Summer Olympics – Qualification_header_cell_5_0_2"
},
{
"operation": "Highlighting deleted",
"time": "687.679",
"value": "Badminton at the 2016 Summer Olympics – Qualification_cell_5_1_2"
},
{
"operation": "Highlighting deleted",
"time": "687.686",
"value": "Badminton at the 2016 Summer Olympics – Qualification_section_2"
},
{
"operation": "Highlighting deleted",
"time": "687.688",
"value": "Badminton at the 2016 Summer Olympics – Qualification_section_7"
},
{
"operation": "Highlighting deleted",
"time": "687.689",
"value": "Badminton at the 2016 Summer Olympics – Qualification_header_cell_5_0_2"
},
{
"operation": "Highlighting",
"time": "693.171",
"value": "Badminton at the 2016 Summer Olympics – Qualification_cell_5_1_2"
},
{
"operation": "Highlighting",
"time": "693.174",
"value": "Badminton at the 2016 Summer Olympics – Qualification_section_2"
},
{
"operation": "Highlighting",
"time": "693.174",
"value": "Badminton at the 2016 Summer Olympics – Qualification_section_7"
},
{
"operation": "Highlighting",
"time": "693.175",
"value": "Badminton at the 2016 Summer Olympics – Qualification_header_cell_5_0_2"
},
{
"operation": "Highlighting",
"time": "694.669",
"value": "Badminton at the 2016 Summer Olympics – Qualification_cell_5_1_3"
},
{
"operation": "Highlighting deleted",
"time": "694.855",
"value": "Badminton at the 2016 Summer Olympics – Qualification_cell_5_1_3"
},
{
"operation": "Highlighting",
"time": "696.95",
"value": "Badminton at the 2016 Summer Olympics – Qualification_cell_5_1_3"
},
{
"operation": "Highlighting",
"time": "697.595",
"value": "Badminton at the 2016 Summer Olympics – Qualification_cell_5_1_4"
},
{
"operation": "Highlighting",
"time": "697.599",
"value": "Badminton at the 2016 Summer Olympics – Qualification_header_cell_5_0_4"
},
{
"operation": "Highlighting",
"time": "699.136",
"value": "Badminton at the 2016 Summer Olympics – Qualification_cell_5_1_0"
},
{
"operation": "Highlighting",
"time": "699.139",
"value": "Badminton at the 2016 Summer Olympics – Qualification_header_cell_5_0_0"
},
{
"operation": "Highlighting deleted",
"time": "700.182",
"value": "Badminton at the 2016 Summer Olympics – Qualification_cell_5_1_0"
},
{
"operation": "Highlighting deleted",
"time": "700.185",
"value": "Badminton at the 2016 Summer Olympics – Qualification_header_cell_5_0_0"
},
{
"operation": "Highlighting",
"time": "700.767",
"value": "Badminton at the 2016 Summer Olympics – Qualification_cell_5_1_1"
},
{
"operation": "Highlighting",
"time": "700.772",
"value": "Badminton at the 2016 Summer Olympics – Qualification_header_cell_5_0_1"
},
{
"operation": "Highlighting",
"time": "713.204",
"value": "Badminton at the 2016 Summer Olympics – Qualification_sentence_1"
},
{
"operation": "Highlighting",
"time": "756.43",
"value": "Badminton at the 2016 Summer Olympics – Qualification_sentence_1"
},
{
"operation": "finish",
"time": "815.462",
"value": "finish"
}
]
}
|
exfever_3132
|
Tatum O'Neal appeared in guest roles in a series that premiered and was set and filmed in the most densely populated major city in the United States.
|
{
"evidence": [
"Tatum O'Neal",
"New York City"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTatum O'Neal appeared in guest roles in a series that premiered and was set and filmed in the most densely populated major city in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tatum O'Neal",
"New York City"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3853,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
74221
|
Built in 1994, the Bean Cemetery occupies a roughly trapezoidal plot of 2.45 acres (0.99 ha).
|
{
"evidence": [
"Bean Cemetery_sentence_3",
"Bean Cemetery_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBuilt in 1994, the Bean Cemetery occupies a roughly trapezoidal plot of 2.45 acres (0.99 ha).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bean Cemetery_sentence_3",
"Bean Cemetery_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3854,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.939",
"value": "?search="
},
{
"operation": "search",
"time": "10.792",
"value": "Bean Cemetery"
},
{
"operation": "Now on",
"time": "11.329",
"value": "Bean Cemetery"
},
{
"operation": "Highlighting",
"time": "129.508",
"value": "Bean Cemetery_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "130.987",
"value": "Bean Cemetery_sentence_0"
},
{
"operation": "Highlighting",
"time": "164.251",
"value": "Bean Cemetery_sentence_3"
},
{
"operation": "Highlighting",
"time": "177.619",
"value": "Bean Cemetery_sentence_5"
},
{
"operation": "finish",
"time": "184.044",
"value": "finish"
}
]
}
|
85767
|
Jansson was educated in physics and sociology at the Örebro University (state university in Örebro, Sweden), before he started a career in ICA, where he held a range of junior positions from 1983 to 1998, and he then worked as CEO of Catena until 1999 when he was hired as CEO of Fazer Group.
|
{
"evidence": [
"Mats Jansson_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJansson was educated in physics and sociology at the Örebro University (state university in Örebro, Sweden), before he started a career in ICA, where he held a range of junior positions from 1983 to 1998, and he then worked as CEO of Catena until 1999 when he was hired as CEO of Fazer Group.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mats Jansson_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3855,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.358",
"value": "?search="
},
{
"operation": "search",
"time": "19.79",
"value": "Contains...Jansson was educated in physics and sociology at the Örebro University"
},
{
"operation": "Now on",
"time": "20.782",
"value": "undefined"
},
{
"operation": "search",
"time": "41.627",
"value": "Contains...CEO of Catena"
},
{
"operation": "Now on",
"time": "42.423",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "47.012",
"value": "Mats Jansson"
},
{
"operation": "Now on",
"time": "47.813",
"value": "Mats Jansson"
},
{
"operation": "Highlighting",
"time": "52.541",
"value": "Mats Jansson_sentence_3"
},
{
"operation": "finish",
"time": "63.993",
"value": "finish"
}
]
}
|
83118
|
Nils Winter is a German long jumper who achieved his personal best jump during the 2004 Summer Olympics.
|
{
"evidence": [
"Nils Winter_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNils Winter is a German long jumper who achieved his personal best jump during the 2004 Summer Olympics.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nils Winter_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3856,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.827",
"value": "?search="
},
{
"operation": "search",
"time": "7.397",
"value": "Nils Winter"
},
{
"operation": "Now on",
"time": "8.116",
"value": "Nils Winter"
},
{
"operation": "Highlighting",
"time": "10.287",
"value": "Nils Winter_sentence_0"
},
{
"operation": "Highlighting",
"time": "19.002",
"value": "Nils Winter_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "22.691",
"value": "Nils Winter_sentence_0"
},
{
"operation": "finish",
"time": "28.165",
"value": "finish"
}
]
}
|
exfever_7216
|
Lindsay Dee Lohan is an actress who appeared in a 2004 American teen musical film with a genre in which songs by the characters are interwoven into the narrative.
|
{
"evidence": [
"Lindsay Lohan",
"Confessions of a Teenage Drama Queen",
"Musical film"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLindsay Dee Lohan is an actress who appeared in a 2004 American teen musical film with a genre in which songs by the characters are interwoven into the narrative.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lindsay Lohan",
"Confessions of a Teenage Drama Queen",
"Musical film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3857,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3381
|
Avatar is a historical adventure television series created by Jonathan E. Steinberg and Robert Levine for an American premium cable and satellite television network owned by Lions Gate Entertainment.
|
{
"evidence": [
"Black Sails (TV series)",
"Starz"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAvatar is a historical adventure television series created by Jonathan E. Steinberg and Robert Levine for an American premium cable and satellite television network owned by Lions Gate Entertainment.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Black Sails (TV series)",
"Starz"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3858,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39013
|
Steven Nitah was born in Canada, his profession is politics he ran in the general election which he was supported by Don Morin and vied to be elected again in 2003.
|
{
"evidence": [
"Steven Nitah_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSteven Nitah was born in Canada, his profession is politics he ran in the general election which he was supported by Don Morin and vied to be elected again in 2003.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Steven Nitah_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3859,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.864",
"value": "?search="
},
{
"operation": "search",
"time": "13.985",
"value": "Steven Nitah"
},
{
"operation": "Now on",
"time": "14.8",
"value": "Steven Nitah"
},
{
"operation": "Highlighting",
"time": "50.561",
"value": "Steven Nitah_sentence_0"
},
{
"operation": "Highlighting",
"time": "70.767",
"value": "Steven Nitah_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "74.086",
"value": "Steven Nitah_sentence_2"
},
{
"operation": "search",
"time": "135.312",
"value": "politicians born in canada"
},
{
"operation": "Now on",
"time": "138.726",
"value": "undefined"
},
{
"operation": "finish",
"time": "184.451",
"value": "finish"
}
]
}
|
exfever_5300
|
Watchmen was the third film by Zack Snyder.
|
{
"evidence": [
"Watchmen",
"Charlton Comics",
"Line (comics)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWatchmen was the third film by Zack Snyder.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Watchmen",
"Charlton Comics",
"Line (comics)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3860,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
6374
|
The hybrid gasoline-electric compact MPV Toyota Prius V is designed like a compact-based minivan retaining the triangular silhouette symbolic of Prius.
|
{
"evidence": [
"Toyota Prius V_sentence_0",
"Toyota Prius V_sentence_6",
"Toyota Prius V_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe hybrid gasoline-electric compact MPV Toyota Prius V is designed like a compact-based minivan retaining the triangular silhouette symbolic of Prius.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Toyota Prius V_sentence_0",
"Toyota Prius V_sentence_6",
"Toyota Prius V_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3861,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.56",
"value": "?search="
},
{
"operation": "search",
"time": "23.061",
"value": "Toyota Prius V"
},
{
"operation": "Now on",
"time": "24.006",
"value": "Toyota Prius V"
},
{
"operation": "search",
"time": "74.145",
"value": "hybrid gasoline-electric compact MPV Toyota Prius V"
},
{
"operation": "Now on",
"time": "75.421",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "81.707",
"value": "Toyota Prius V"
},
{
"operation": "Now on",
"time": "82.042",
"value": "Toyota Prius V"
},
{
"operation": "Page search",
"time": "86.825",
"value": "hybrid"
},
{
"operation": "Highlighting",
"time": "91.488",
"value": "Toyota Prius V_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "91.757",
"value": "Toyota Prius V_sentence_0"
},
{
"operation": "Highlighting",
"time": "92.216",
"value": "Toyota Prius V_sentence_0"
},
{
"operation": "Highlighting",
"time": "114.486",
"value": "Toyota Prius V_sentence_6"
},
{
"operation": "Highlighting",
"time": "138.179",
"value": "Toyota Prius V_sentence_7"
},
{
"operation": "finish",
"time": "167.621",
"value": "finish"
}
]
}
|
exfever_3880
|
Logan is a 2017 Canadian superhero film starring Hugh Jackman a fictional character appearing in American comic books published by Marvel Comics.
|
{
"evidence": [
"Logan (film)",
"Wolverine (character)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLogan is a 2017 Canadian superhero film starring Hugh Jackman a fictional character appearing in American comic books published by Marvel Comics.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Logan (film)",
"Wolverine (character)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3862,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3740
|
"Umbrella" topped the charts in 13 countries and remained at No. 1 in the UK for 10 consecutive weeks.
|
{
"evidence": [
"Rihanna",
"Bridgetown"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"Umbrella\" topped the charts in 13 countries and remained at No. 1 in the UK for 10 consecutive weeks.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rihanna",
"Bridgetown"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3863,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39113
|
The Tabascans took Villahermosa upriver after bombarding the city on their second attempt; the Tabascans formed guerilla groups which, when matched with the inhospitable drug lords, forced U.S. Military to withdraw.
|
{
"evidence": [
"Tabasco_sentence_131",
"Tabasco_sentence_132",
"Tabasco_sentence_133"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Tabascans took Villahermosa upriver after bombarding the city on their second attempt; the Tabascans formed guerilla groups which, when matched with the inhospitable drug lords, forced U.S. Military to withdraw.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tabasco_sentence_131",
"Tabasco_sentence_132",
"Tabasco_sentence_133"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3864,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.705",
"value": "?search="
},
{
"operation": "search",
"time": "15.578",
"value": "Tabascans"
},
{
"operation": "Now on",
"time": "16.327",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "24.377",
"value": "Tabasco"
},
{
"operation": "Now on",
"time": "25.246",
"value": "Tabasco"
},
{
"operation": "Page search",
"time": "36.683",
"value": "villhermosa"
},
{
"operation": "Page search",
"time": "60.094",
"value": "upriver"
},
{
"operation": "Highlighting",
"time": "94.342",
"value": "Tabasco_sentence_131"
},
{
"operation": "Highlighting",
"time": "102.275",
"value": "Tabasco_sentence_132"
},
{
"operation": "Highlighting",
"time": "113.347",
"value": "Tabasco_sentence_133"
},
{
"operation": "finish",
"time": "153.284",
"value": "finish"
}
]
}
|
62357
|
Qingtian County is a 2,493 km² county in Lishui city, southeastern Hubei Province, in the People's Republic of China with a total population of 361,062 as of 2010.
|
{
"evidence": [
"Qingtian County_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nQingtian County is a 2,493 km² county in Lishui city, southeastern Hubei Province, in the People's Republic of China with a total population of 361,062 as of 2010.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Qingtian County_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3865,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.719",
"value": "?search="
},
{
"operation": "search",
"time": "21.975",
"value": "Qingtian County"
},
{
"operation": "Now on",
"time": "22.743",
"value": "Qingtian County"
},
{
"operation": "Highlighting",
"time": "62.114",
"value": "Qingtian County_cell_0_5_1"
},
{
"operation": "Highlighting",
"time": "62.125",
"value": "Qingtian County_header_cell_0_5_0"
},
{
"operation": "Highlighting",
"time": "62.125",
"value": "Qingtian County_header_cell_0_4_0"
},
{
"operation": "Page search",
"time": "82.376",
"value": "Lishui"
},
{
"operation": "Highlighting",
"time": "90.517",
"value": "Qingtian County_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "90.527",
"value": "Qingtian County_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "90.528",
"value": "Qingtian County_header_cell_0_0_0"
},
{
"operation": "Page search",
"time": "110.899",
"value": "Hubei"
},
{
"operation": "Page search",
"time": "124.563",
"value": "province"
},
{
"operation": "Highlighting deleted",
"time": "194.599",
"value": "Qingtian County_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "194.609",
"value": "Qingtian County_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "194.61",
"value": "Qingtian County_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "195.182",
"value": "Qingtian County_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "195.192",
"value": "Qingtian County_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "195.193",
"value": "Qingtian County_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "200.064",
"value": "Qingtian County_sentence_1"
},
{
"operation": "finish",
"time": "216.995",
"value": "finish"
}
]
}
|
65423
|
Riemannian geometry, a concept first generalized in the 19th century by Bernhard Riemann, deals with geometric shapes that are defined in nature.
|
{
"evidence": [
"Riemannian geometry_sentence_7"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRiemannian geometry, a concept first generalized in the 19th century by Bernhard Riemann, deals with geometric shapes that are defined in nature.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Riemannian geometry_sentence_7"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3866,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.791",
"value": "undefined"
},
{
"operation": "search",
"time": "7.412",
"value": "Riemannian geometry"
},
{
"operation": "Now on",
"time": "8.627",
"value": "Riemannian geometry"
},
{
"operation": "Highlighting",
"time": "78.888",
"value": "Riemannian geometry_sentence_7"
},
{
"operation": "Page search",
"time": "156.412",
"value": "nature"
},
{
"operation": "Page search",
"time": "168.846",
"value": "geometric"
},
{
"operation": "Page search",
"time": "183.884",
"value": "shapes"
},
{
"operation": "search",
"time": "304.614",
"value": "geometric shapes that are defined in nature"
},
{
"operation": "Now on",
"time": "307.475",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "329.505",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "329.734",
"value": "Riemannian geometry"
},
{
"operation": "finish",
"time": "352.71",
"value": "finish"
}
]
}
|
40400
|
Andrea Anderson had to return her medal along with the rest of the after Marion Jones (a Belizean-American former world champion track and field athlete) was disqualified, and in 2010 the court of Arbitration for Sport ruled in favor of the other teammates and returned the medals.
|
{
"evidence": [
"Andrea Anderson_sentence_2",
"Andrea Anderson_sentence_3",
"Marion Jones_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAndrea Anderson had to return her medal along with the rest of the after Marion Jones (a Belizean-American former world champion track and field athlete) was disqualified, and in 2010 the court of Arbitration for Sport ruled in favor of the other teammates and returned the medals.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Andrea Anderson_sentence_2",
"Andrea Anderson_sentence_3",
"Marion Jones_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3867,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.73",
"value": "?search="
},
{
"operation": "search",
"time": "33.38",
"value": "Andrea Anderson"
},
{
"operation": "Now on",
"time": "35.322",
"value": "Andrea Anderson"
},
{
"operation": "Highlighting",
"time": "42.819",
"value": "Andrea Anderson_sentence_2"
},
{
"operation": "Highlighting",
"time": "51.998",
"value": "Andrea Anderson_sentence_3"
},
{
"operation": "hyperlink",
"time": "56.022",
"value": "Marion Jones"
},
{
"operation": "Now on",
"time": "58.928",
"value": "Marion Jones"
},
{
"operation": "Highlighting",
"time": "60.59",
"value": "Marion Jones_sentence_1"
},
{
"operation": "finish",
"time": "116.795",
"value": "finish"
}
]
}
|
51294
|
UEFA Euro 1988 squads had 20 players, the person in first position was Sjaak Troost (born 28 August 1959).
|
{
"evidence": [
"UEFA Euro 1988_sentence_78"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nUEFA Euro 1988 squads had 20 players, the person in first position was Sjaak Troost (born 28 August 1959).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"UEFA Euro 1988_sentence_78"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3868,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.1",
"value": "?search="
},
{
"operation": "search",
"time": "17.025",
"value": "UEFA Euro 1988"
},
{
"operation": "Now on",
"time": "23.624",
"value": "UEFA Euro 1988"
},
{
"operation": "Highlighting",
"time": "50.162",
"value": "UEFA Euro 1988_sentence_78"
},
{
"operation": "hyperlink",
"time": "139.88",
"value": "UEFA Euro 1988 Final"
},
{
"operation": "Now on",
"time": "140.945",
"value": "UEFA Euro 1988 Final"
},
{
"operation": "search",
"time": "168.576",
"value": "Sjaak Troost"
},
{
"operation": "Now on",
"time": "171.481",
"value": "Sjaak Troost"
},
{
"operation": "hyperlink",
"time": "183.816",
"value": "1988"
},
{
"operation": "Now on",
"time": "185.052",
"value": "UEFA Euro 1988"
},
{
"operation": "back-button-clicked",
"time": "197.033",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "197.147",
"value": "Sjaak Troost"
},
{
"operation": "hyperlink",
"time": "198.943",
"value": "UEFA European Championship"
},
{
"operation": "Now on",
"time": "203.676",
"value": "UEFA European Championship"
},
{
"operation": "back-button-clicked",
"time": "207.624",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "207.706",
"value": "Sjaak Troost"
},
{
"operation": "hyperlink",
"time": "220.447",
"value": "UEFA Euro 1988"
},
{
"operation": "Now on",
"time": "221.11",
"value": "UEFA Euro 1988"
},
{
"operation": "finish",
"time": "311.316",
"value": "finish"
}
]
}
|
57343
|
James Kim (August 9, 1971 – December 4, 2006 due to hypothermia) was an American radio personality and technology analyst from Oregon whose disappearance and death became a subject of a brief but intense period of news coverage in December 2006.
|
{
"evidence": [
"James Kim_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJames Kim (August 9, 1971 – December 4, 2006 due to hypothermia) was an American radio personality and technology analyst from Oregon whose disappearance and death became a subject of a brief but intense period of news coverage in December 2006.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"James Kim_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3869,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.204",
"value": "?search="
},
{
"operation": "search",
"time": "52.327",
"value": "James Kim"
},
{
"operation": "Now on",
"time": "53.245",
"value": "James Kim"
},
{
"operation": "Highlighting",
"time": "116.517",
"value": "James Kim_sentence_1"
},
{
"operation": "finish",
"time": "1058.425",
"value": "finish"
}
]
}
|
exfever_1066
|
Jonah Hill Feldstein was a successful American actor from 2014 to 2015.
|
{
"evidence": [
"Jonah Hill",
"Forgetting Sarah Marshall"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJonah Hill Feldstein was a successful American actor from 2014 to 2015.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jonah Hill",
"Forgetting Sarah Marshall"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3870,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
42530
|
De'Vondre Campbell plays for the Arizona Cardinals and Blake Cashman plays for the New York Jets.
|
{
"evidence": [
"De'Vondre Campbell_sentence_0",
"Blake Cashman_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDe'Vondre Campbell plays for the Arizona Cardinals and Blake Cashman plays for the New York Jets.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"De'Vondre Campbell_sentence_0",
"Blake Cashman_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3871,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.712",
"value": "?search="
},
{
"operation": "search",
"time": "6.357",
"value": "De'Vondre Campbell"
},
{
"operation": "Now on",
"time": "7.554",
"value": "De'Vondre Campbell"
},
{
"operation": "Highlighting",
"time": "12.431",
"value": "De'Vondre Campbell_sentence_0"
},
{
"operation": "search",
"time": "19.22",
"value": "blake cashman"
},
{
"operation": "Now on",
"time": "19.917",
"value": "Blake Cashman"
},
{
"operation": "Highlighting",
"time": "21.847",
"value": "Blake Cashman_sentence_0"
},
{
"operation": "finish",
"time": "27.373",
"value": "finish"
}
]
}
|
71951
|
The IWRG 6th Anniversary Show saw El Príncipe de Fuego defeat La Sombra in a Best two-out-of-three-falls singles match, and while the match between Capitán Sangre and Guerrillero, and Meteoro and Star Boy in a Best two-out-of-three-falls tag team match was postponed.
|
{
"evidence": [
"IWRG 6th Anniversary Show_sentence_23"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe IWRG 6th Anniversary Show saw El Príncipe de Fuego defeat La Sombra in a Best two-out-of-three-falls singles match, and while the match between Capitán Sangre and Guerrillero, and Meteoro and Star Boy in a Best two-out-of-three-falls tag team match was postponed.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"IWRG 6th Anniversary Show_sentence_23"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3872,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.695",
"value": "?search="
},
{
"operation": "search",
"time": "48.821",
"value": "IWRG 6th Anniversary Show"
},
{
"operation": "Now on",
"time": "49.867",
"value": "IWRG 6th Anniversary Show"
},
{
"operation": "Page search",
"time": "53.9",
"value": "fuego"
},
{
"operation": "Highlighting",
"time": "62.473",
"value": "IWRG 6th Anniversary Show_sentence_22"
},
{
"operation": "Highlighting",
"time": "87.701",
"value": "IWRG 6th Anniversary Show_sentence_23"
},
{
"operation": "Highlighting deleted",
"time": "90.69",
"value": "IWRG 6th Anniversary Show_sentence_22"
},
{
"operation": "finish",
"time": "99.814",
"value": "finish"
}
]
}
|
exfever_5971
|
Leslie Groves was a United States Army Corps of Engineers officer who oversaw the construction of the Pentagon and directed the Manhattan Project, a top secret research project that developed the atomic bomb during a global war that lasted from 1939 to 1945.
|
{
"evidence": [
"Leslie Groves",
"World War II"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLeslie Groves was a United States Army Corps of Engineers officer who oversaw the construction of the Pentagon and directed the Manhattan Project, a top secret research project that developed the atomic bomb during a global war that lasted from 1939 to 1945.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leslie Groves",
"World War II"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3873,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3775
|
Alpha House is a political satire streaming series starring an American actor and singer who has also starred in films, including a black thriller crime film produced and directed by Joel and Ethan Coen.
|
{
"evidence": [
"Alpha House",
"John Goodman",
"The Big Lebowski"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAlpha House is a political satire streaming series starring an American actor and singer who has also starred in films, including a black thriller crime film produced and directed by Joel and Ethan Coen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alpha House",
"John Goodman",
"The Big Lebowski"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3874,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
43431
|
MG George E. Pickett, a career United States Army officer who was promoted as major general in the Confederate States Army during the American Civil War, commandeered five brigades in the Department of North Carolina and Southern Virginia.
|
{
"evidence": [
"George Pickett_sentence_2",
"George Pickett_sentence_135"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMG George E. Pickett, a career United States Army officer who was promoted as major general in the Confederate States Army during the American Civil War, commandeered five brigades in the Department of North Carolina and Southern Virginia.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"George Pickett_sentence_2",
"George Pickett_sentence_135"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3875,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.394",
"value": "?search="
},
{
"operation": "search",
"time": "13.658",
"value": "George E. Pickett"
},
{
"operation": "Now on",
"time": "15.231",
"value": "George Pickett"
},
{
"operation": "Highlighting",
"time": "37.948",
"value": "George Pickett_sentence_2"
},
{
"operation": "hyperlink",
"time": "105.062",
"value": "3.4 North Carolina George Pickett_section_6"
},
{
"operation": "Highlighting",
"time": "112.253",
"value": "George Pickett_sentence_135"
},
{
"operation": "finish",
"time": "154.953",
"value": "finish"
}
]
}
|
exfever_3168
|
Nick Jonas formed the Jonas Brothers with his older brothers, including a member of the band that signed with an American record label.
|
{
"evidence": [
"Nick Jonas",
"Joe Jonas",
"Hollywood Records"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNick Jonas formed the Jonas Brothers with his older brothers, including a member of the band that signed with an American record label.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Nick Jonas",
"Joe Jonas",
"Hollywood Records"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3876,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
69636
|
2011 South Korean Figure Skating Championships, organized by the Korean Skating Union (founded on November 16,1999), were the 65th edition of those championships held.
|
{
"evidence": [
"2011 South Korean Figure Skating Championships_sentence_2",
"2011 South Korean Figure Skating Championships_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n2011 South Korean Figure Skating Championships, organized by the Korean Skating Union (founded on November 16,1999), were the 65th edition of those championships held.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"2011 South Korean Figure Skating Championships_sentence_2",
"2011 South Korean Figure Skating Championships_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3877,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "3.758",
"value": "?search="
},
{
"operation": "search",
"time": "46.099",
"value": "2011 South Korean Figure Skating Championships"
},
{
"operation": "Now on",
"time": "48.554",
"value": "2011 South Korean Figure Skating Championships"
},
{
"operation": "Highlighting",
"time": "63.405",
"value": "2011 South Korean Figure Skating Championships_sentence_2"
},
{
"operation": "hyperlink",
"time": "159.056",
"value": "South Korean Figure Skating Championships"
},
{
"operation": "Now on",
"time": "172.155",
"value": "Index.php?title=Korean Skating Union&action=edit&redlink=1"
},
{
"operation": "back-button-clicked",
"time": "175.912",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "182.718",
"value": "2011 South Korean Figure Skating Championships"
},
{
"operation": "Highlighting deleted",
"time": "186.383",
"value": "2011 South Korean Figure Skating Championships_sentence_2"
},
{
"operation": "Highlighting",
"time": "187.61",
"value": "2011 South Korean Figure Skating Championships_sentence_2"
},
{
"operation": "search",
"time": "190.824",
"value": "Korean Skating Union"
},
{
"operation": "Now on",
"time": "194.337",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "234.678",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "234.812",
"value": "2011 South Korean Figure Skating Championships"
},
{
"operation": "Highlighting",
"time": "445.214",
"value": "2011 South Korean Figure Skating Championships_sentence_1"
},
{
"operation": "search",
"time": "469.45",
"value": "Korean Skating Union founded"
},
{
"operation": "Now on",
"time": "472.057",
"value": "undefined"
},
{
"operation": "finish",
"time": "483.858",
"value": "finish"
}
]
}
|
2415
|
1948 Palestinian exodus (also known as the Nakba) had over 800,000 refugees from 29 Nov. 1947 to 20 July 1949.
|
{
"evidence": [
"1948 Palestinian exodus_sentence_0",
"1948 Palestinian exodus_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n1948 Palestinian exodus (also known as the Nakba) had over 800,000 refugees from 29 Nov. 1947 to 20 July 1949.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"1948 Palestinian exodus_sentence_0",
"1948 Palestinian exodus_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3878,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "14.721",
"value": "1948 Palestinian exodus"
},
{
"operation": "Highlighting",
"time": "40.231",
"value": "1948 Palestinian exodus_sentence_0"
},
{
"operation": "Highlighting",
"time": "53.407",
"value": "1948 Palestinian exodus_sentence_2"
},
{
"operation": "finish",
"time": "67.819",
"value": "finish"
}
]
}
|
exfever_2720
|
Jennifer Grey is an American actress who subsequently earned worldwide fame for starring as Frances "Baby" Houseman in a 1987 American dance film written by Eleanor Bergstein that involves intimate love stories recorded in visual media for broadcast in theatres or on television.
|
{
"evidence": [
"Jennifer Grey",
"Dirty Dancing",
"Romance film"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJennifer Grey is an American actress who subsequently earned worldwide fame for starring as Frances \"Baby\" Houseman in a 1987 American dance film written by Eleanor Bergstein that involves intimate love stories recorded in visual media for broadcast in theatres or on television.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jennifer Grey",
"Dirty Dancing",
"Romance film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3879,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
35659
|
Teams Olimpija Ljubljana and Domžale qualified for the 2017–18 Europa League first qualifying round.
|
{
"evidence": [
"Olimpija Ljubljana_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTeams Olimpija Ljubljana and Domžale qualified for the 2017–18 Europa League first qualifying round.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Olimpija Ljubljana_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3880,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.742",
"value": "?search="
},
{
"operation": "search",
"time": "12.796",
"value": "Contains...2017–18 Europa League"
},
{
"operation": "Now on",
"time": "14.845",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "24.03",
"value": "2018 UEFA Europa League Final"
},
{
"operation": "Now on",
"time": "24.819",
"value": "2018 UEFA Europa League Final"
},
{
"operation": "search",
"time": "65.056",
"value": "Contains...2017–18 Europa League first qualifying round"
},
{
"operation": "Now on",
"time": "67.052",
"value": "undefined"
},
{
"operation": "search",
"time": "87.304",
"value": "Olimpija Ljubljana"
},
{
"operation": "Now on",
"time": "87.98",
"value": "Olimpija Ljubljana"
},
{
"operation": "search",
"time": "111.045",
"value": "Contains...2017–18 Europa League"
},
{
"operation": "Now on",
"time": "111.684",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "131.106",
"value": "2018 UEFA Europa League Final"
},
{
"operation": "Now on",
"time": "131.681",
"value": "2018 UEFA Europa League Final"
},
{
"operation": "search",
"time": "148.324",
"value": "Domžale"
},
{
"operation": "Now on",
"time": "149.124",
"value": "Domžale"
},
{
"operation": "search",
"time": "175.208",
"value": "Contains...Teams Olimpija Ljubljana and Domžale"
},
{
"operation": "Now on",
"time": "176.613",
"value": "undefined"
},
{
"operation": "search",
"time": "199.635",
"value": "Contains...qualified for the 2017–18 Europa League"
},
{
"operation": "Now on",
"time": "201.69",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "208.426",
"value": "2016–17 FA Cup"
},
{
"operation": "Now on",
"time": "209.185",
"value": "2016–17 FA Cup"
},
{
"operation": "search",
"time": "233.592",
"value": "Olimpija Ljubljana"
},
{
"operation": "Now on",
"time": "234.95",
"value": "Olimpija Ljubljana"
},
{
"operation": "Highlighting",
"time": "249.281",
"value": "Olimpija Ljubljana_sentence_0"
},
{
"operation": "finish",
"time": "255.546",
"value": "finish"
}
]
}
|
exfever_6174
|
Elsa Lafuente Medianu appeared in several films, including a 2009 American neo-noir action crime thriller film directed by an English film director.?
|
{
"evidence": [
"Elsa Pataky",
"Give 'Em Hell, Malone",
"Russell Mulcahy"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nElsa Lafuente Medianu appeared in several films, including a 2009 American neo-noir action crime thriller film directed by an English film director.?\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Elsa Pataky",
"Give 'Em Hell, Malone",
"Russell Mulcahy"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3881,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5269
|
The Great British Sewing Bee is a reality show that is a spin-off of the format of a television baking competition whose first four series were broadcast on a British free-to-air television network.
|
{
"evidence": [
"The Great British Sewing Bee",
"The Great British Bake Off",
"BBC Two"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Great British Sewing Bee is a reality show that is a spin-off of the format of a television baking competition whose first four series were broadcast on a British free-to-air television network.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Great British Sewing Bee",
"The Great British Bake Off",
"BBC Two"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3882,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32193
|
Islands District Council has more than ten Constituency one of which involves Josephine Tsang Sau-ho who scored a little over 50% of the votes.
|
{
"evidence": [
"Islands District Council_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIslands District Council has more than ten Constituency one of which involves Josephine Tsang Sau-ho who scored a little over 50% of the votes.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Islands District Council_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3883,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.161",
"value": "?search="
},
{
"operation": "search",
"time": "16",
"value": "Islands District Council"
},
{
"operation": "Now on",
"time": "17.138",
"value": "Islands District Council"
},
{
"operation": "Highlighting",
"time": "24.698",
"value": "Islands District Council_sentence_2"
},
{
"operation": "finish",
"time": "84.049",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.